State of Crypto Withdrawal Freezes: Q3 2026 Data Report (28,500+ events tracked across 7 exchanges)

This article contains affiliate links. Yieldo may earn a commission at no extra cost to you. This does not affect the data — the Withdrawal Reliability Ranking is derived from an open event log you can independently verify.

Q3 2026 opens with roughly 28.9% of crypto withdrawal routes frozen and 56.2% of deposit routes disabled across seven supported exchanges — data that exchanges themselves never publish, tracked from a live event log of 28,500+ freeze and restore events. Closed routes are not edge cases. They are the structural backdrop of every cross-exchange transfer, and the numbers behind that backdrop are what this report is here to document.

This is Yieldo's first quarterly withdrawal-freeze report. It aggregates the 25 June 2026 opening snapshot with the first seven days of Q3 event-log data and lays out the methodology that will drive the full Q4 2026 retrospective in October. If you want the live state of any route, the tracker at /fees/network-status is refreshed every 30 minutes; if you need the emergency playbook when your own route is closed, jump to our withdrawal-network-disabled how-to.

Executive Summary — Q3 2026 Baseline in Numbers

The Q3 2026 Opening Snapshot at a Glance

On the 25 June 2026 reference snapshot, out of 23,118 coin-network route pairs covered across seven exchanges: 6,681 (28.9%) withdrawal routes and 12,990 (56.2%) deposit routes were flagged off. Roughly 40-45% of routes were fully operational — both withdraw and deposit switched on. Between the snapshot and Q3 opening, the ranges held steady: ~28-30% of withdraw routes and ~55-60% of deposit routes disabled at any random sampling moment.

That gap between withdraw and deposit is not an accident. It is the structural fingerprint of a deposit pipeline that has more failure surfaces than a withdrawal pipeline, and it is the single most useful lens through which to read this report.

What the Numbers Mean (28.9% Withdrawals / 56.2% Deposits Frozen)

If you picked one CEX at random, one coin at random, and one network at random from the tracker at Q3 opening, roughly three in ten times the withdrawal button would be off. That is not a market-wide crisis. That is Tuesday. Closed routes are not edge cases — they are the structural backdrop of cross-exchange transfers, and reading them as edge cases is what causes fund-movers to stall mid-transfer with an urgent asset and no way out.

The deposit share matters even more for a different reason: exchanges bias toward closing the inbound door first. Deposits are the pipeline that has to survive a chain indexer, a memo parser, a hot-wallet crediting stage, and an internal balance update — four stages, any one of which can fail. Withdrawals only need a signing service and a broadcaster. When something is being investigated, the deposit switch gets flipped as a precaution, and it usually stays off longer than the withdrawal switch on the same route.

Who Freezes Most — and Who Freezes Least — This Quarter

We publish a Withdrawal Reliability Ranking further down in this report. At Q3 opening, based on nine days of event-log data plus structural facts from the last twelve months of public record:

1. Bybit — lowest freeze frequency, most maintenance windows pre-announced. Q3 reliability winner. Sign up at Bybit if the reliability tier is what you're optimizing for.
2. Bitget — no headline structural sunsets, upgrade-related freezes only, short durations. Secondary reliability winner. Onboard at Bitget.
3. KuCoin — mid-tier freeze frequency, meaningful US-access limitation post-CFTC March 2024.
4. MEXC — deepest catalog produces highest raw event count but shortest individual freezes.
5. OKX — high headline suspended share, inflated by 34 permanently disabled OKTC legacy routes.
6. Gate.io — deepest long-tail altcoin catalog, highest raw suspended share by design, not necessarily reliability signal.

This is a provisional ranking. The full Withdrawal Reliability Score with all four components — freeze frequency, average duration, flap rate, structural disabled share — publishes in the Q4 2026 report. We show our work in the Methodology section so you can reweight it yourself.

Methodology: How We Track Withdrawal Freezes

The Data Source — network_availability_events Log

The report is built on two data structures inside the Yieldo tracker: (1) the withdrawal_fees snapshot table, which stores the current withdraw_enabled and deposit_enabled flag on every (exchange, coin, network) triplet across the seven CEX we cover; and (2) the network_availability_events log, which records every observed transition of those flags as a row with recorded_at timestamp, exchange, coin, network, and event_type ∈ {withdraw_froze, withdraw_restored, deposit_froze, deposit_restored}.

The snapshot table is refreshed on a 30-minute cadence via the same REST endpoints that each exchange's wallet team uses to render deposit and withdraw modals in production apps. The event log came online on 2026-06-28 and has accumulated roughly 28,500 events in the first nine days — a cadence of approximately 3,166 events per day, or 132 per hour, across all six data-eligible CEX × every coin × every network route.

You can verify these numbers from /fees/network-status.json (live feed) and via the CSV dump on request. The full aggregation logic lives at /about/data.

How Freeze and Restore Events Are Classified

An event is written into the log only when the snapshot table observes a genuine flip on a route that already existed in the previous snapshot. That guard-rail matters because our aggregator runs a markStaleAsUnavailable() routine on the snapshot table: any route that has not refreshed within a two-hour window is auto-marked withdraw_enabled = false on the snapshot. Those stale-flips are excluded from the event log by design, via a lastWithdrawEventWasFroze check inside the write path. Snapshot numbers (28.9% / 56.2%) may include 1-2% transient false positives from short ingestion gaps; event-log numbers do not.

Event types:

• withdraw_froze — flag flipped from true to false on a live route
• withdraw_restored — flag flipped from false to true on a route previously in a withdraw_froze state
• deposit_froze / deposit_restored — mirror events for the deposit switch on the same triplet

The classification is intentionally atomic: one row per transition, no derived states in the log. Derivations — mean-time-to-recovery, flap rate, average downtime — all compute from paired {froze, restored} events at query time.

Limitations, Reproducibility, and Honesty Framing (9-Day Sample)

We publish this report as an opening snapshot, not as a full quarterly retrospective, and we are explicit about why:

• Nine days of event history at time of writing. The event log starts 2026-06-28. Longitudinal metrics — mean-time-to-recovery, weekly flap rankings, cross-week freeze-frequency trends — require several more weeks. The Q4 2026 report (mid-October) will draw on roughly 15× more history.
• Stale-flip false positives are in the snapshot but not in the event log. Reading a snapshot as ground truth for freeze_frequency will overstate the number by 1-2 percentage points; reading it as a directional headline is fine.
• Not every exchange pre-announces maintenance. Bybit typically does. MEXC and Gate.io often do not. This shifts events between our "planned" and "unplanned" root-cause buckets but does not distort the counting.
• US-jurisdiction access is not uniform. Bybit, MEXC, OKX, and Gate.io restrict US-resident access, KuCoin closed US access after the March 2024 CFTC settlement, and Bitget never served US users. Every exchange remains in scope for the tracker itself, but the "which exchange is more reliable for you" question is jurisdiction-scoped.
• DEXes are out of scope. STON.fi on TON and Jupiter on Solana do not have a withdraw or deposit toggle: users hold funds themselves. Their failure modes — governance-paused smart contracts, chain outages — are structurally different and are covered separately in the DEX escape-hatch section below.

We keep a separate research-study on the staking counterpart of this problem — the way APR tiers hide rate cuts in plain sight — at do exchanges cut APY after listing. Same methodology posture, different vertical.

Snapshot: State of Withdrawals at Q3 2026 Opening

Withdrawals Disabled vs Enabled (Coin-Network Pairs)

The 25 June 2026 reference snapshot marks roughly 28.9% of withdrawal routes off across the seven exchanges we cover. In prose we hold that as ~28-30% at any random moment: the range has been stable across recent weeks, and the Q3 opening reading sits inside it. Closed routes are not edge cases — they are the structural backdrop of cross-exchange transfers.

Reading the widget above: the two large headline numbers are the current suspended-network count and the count of freeze events in the last 30 days. The table below them lists the first eight networks currently in a suspended state, with withdraw_enabled and deposit_enabled both visible so you can see the deposit-side asymmetry route by route. When you plan a transfer, that widget — not the "supported networks" list on the exchange's own coin page — is the source of truth for the next 30 minutes.

Deposits Disabled vs Enabled (The Bigger Half of the Problem)

Deposits are the bigger half of the problem. ~55-60% of deposit routes are off at any given moment, roughly 1.94× the withdrawal-side share. The 56.2% opening reading has been stable within a two-percentage-point band for several weeks.

Why is the deposit switch flipped more often? Because the deposit pipeline has four failure surfaces: (1) the chain indexer that watches the deposit address, (2) the memo parser that credits the correct sub-account for coins like XRP, XLM, TON, and BEP2, (3) the hot-wallet crediting stage that moves funds into house liquidity, and (4) the internal balance update that shows the user their new balance. Any one of the four can fail — and when the wallet team is unsure, they close the inbound door first. Closed deposits do not make headlines. Closed withdrawals do.

For a fund-mover, the practical implication is that you should always confirm the deposit side on the destination exchange before initiating the withdrawal on the source exchange. A route with a green withdraw_enabled on the source and a red deposit_enabled on the destination is not a route. It is a way to have your funds land in a limbo state on-chain.

Most-Frozen Coins at Q3 Opening

The most-frozen coins at Q3 opening are dominated by two categories:

• Long-tail altcoins with narrow liquidity — MEXC and Gate.io each list several thousand tokens where individual routes auto-pause when maker inventory thresholds are hit. Individual freezes are typically hours, not days, but the raw event count is high because the population is large.
• Coins on legacy or sunsetting chains — the 34 OKX OKTC routes account for a disproportionate share of "permanently disabled" flags on OKX. MEXC's ZKSYNC LITE route for ETH and Bitget's BITLAYER route for BTC are the other headline structural sunsets. These are not "freezes" in an operational sense — they are chain sunsets. See the live coins-you-cant-withdraw tracker for the current per-coin frozen-route list.

For the majority-liquidity coins — BTC, ETH, USDT, USDC, SOL, TON, XRP — the freeze picture is much thinner. USDT freezes are almost always on peripheral networks (Polygon, Optimism, ARB, ZKSYNC ERA); TRC20 and ERC20 rails are only ever taken down for major chain events. See cheapest USDT withdrawal and BTC withdrawal for the specific reliable routes.

Most-Frozen Networks at Q3 Opening

The most-frozen networks aggregate into four groups:

• Legacy L1s being sunsetted — OKTC (OKX), ZKSYNC LITE (industry-wide), BITLAYER (Bitget-specific)
• New L2s during rollup upgrade windows — Base, Optimism, Arbitrum, and ZKSYNC ERA each go through short precautionary freezes during rollup version bumps
• Chain-specific memo-required networks — XLM and XRP routes flip more often during exchange indexer changes because of the memo/tag write path
• Deprecated Binance-family rails — BEP2 continues to see gradual deprecation events

Nothing on this list is surprising. What is surprising to first-time readers is how often stable rails — ERC20 and TRC20 for USDT, native BTC for BTC, TON for TON — appear on the frozen list too, whenever an exchange is doing wallet maintenance. Assuming your favourite rail is "always up" is the single most expensive assumption a fund-mover can make.

First-Week Q3 Trends (July 1-7, 2026)

Event Volume Per Exchange (First Week)

The first seven days of Q3 recorded approximately 22,000 events across the six data-eligible CEX (MEXC, Bitget, OKX, Bybit, Gate.io, KuCoin) — an average of roughly 3,150 events per day. That is the run rate. Any headline in the next quarterly report that reads "5,000 events per day" is a signal that something structural changed; anything within a 20% band around 3,150 is noise.

Event volume is heavily skewed by catalog size. MEXC and Gate.io — with deep low-cap altcoin catalogs — produce the majority of the raw event count. Bybit and Bitget — with narrower, quality-curated catalogs — produce the smallest raw counts. Raw event count is not a reliability signal on its own: MEXC producing 8× the events of Bybit is not "8× less reliable" — it is "8× more surface area."

The reliability signal is events / active_routes, and by that measure the ordering is very different from the raw ordering. That normalized ordering is what feeds the Withdrawal Reliability Ranking further below.

Freeze-to-Restore Ratio and Median Downtime

Over the first week of Q3, withdraw_restored events matched withdraw_froze events within a 1:1 ± 5% band — every freeze paired with a restore within the observation window. That is a healthy signal: if the restore-to-freeze ratio were falling below 0.8, we would be looking at a market where routes are getting closed and not reopened, and the aggregate suspended share would trend up.

Median downtime — for the routes that flipped and flipped back within the nine-day window — sat in the single-digit-hours range for planned-maintenance events and in the 24-72 hour range for security-incident-adjacent events. We will publish real medians per exchange with 30× more history in the Q4 report.

Signal vs Noise — Excluding Stale-Data Flips

We have said it twice already but it bears one more mention because it is the single most common misreading of this dataset: the snapshot table and the event log are not the same source. The snapshot table includes stale-flip artifacts that get auto-flipped to withdraw_enabled = false when the ingestion pipeline misses a two-hour refresh window. The event log excludes those artifacts by design. A row in the event log always corresponds to a real transition observed in a fresh snapshot immediately following a fresh snapshot. That is the guardrail that keeps the reliability metrics honest.

If a critic asks "how do I know your Withdrawal Reliability Ranking is not just an artifact of your ingestion pipeline going down at bad times?" — the answer is: because the ranking is computed from the event log, and the event log is guarded against exactly that failure mode. If you want to check us anyway, download the CSV dump and reproduce the calculation. That is what open data is for.

Per-Exchange Breakdown: Withdrawal Reliability at Q3 Opening

MEXC — Highest Route Count, Aggressive Delisting Pattern

MEXC operates one of the deepest low-cap altcoin catalogs in the CEX universe and correspondingly has the highest raw event count in our tracker. Individual freezes on MEXC are typically hours, not days, and the wallet team is aggressive about pruning low-volume routes — which shows up in the log as a steady stream of small events rather than a small number of large events. The MEXC withdrawal fee page breaks down the per-network cost side of the same picture.

MEXC's ZKSYNC LITE route for ETH is a permanent structural sunset, not a freeze — the same class as the OKX OKTC block below. The 9 February 2026 MX Holder Discount suspension is a separate matter — it touched a spot-fee discount program, not the withdrawal engine, and does not appear in this dataset. For a low-friction on-ramp with the widest catalog on the market, MEXC remains a rational choice provided you check the tracker before every large transfer.

Bybit — Q3 Baseline After the 2025 Lazarus Aftermath

Bybit is the Q3 reliability winner in our tracker, both on raw event count normalized by active routes and on the ratio of pre-announced-to-unannounced maintenance windows. The February 2025 Lazarus incident — the $1.5 billion cold-wallet compromise — resulted in a 24-48 hour pause on ETH-related withdrawals, followed by full restoration. Sixteen months later, the incident is what analysts call an evergreen data point: it happened, it was severe, and the post-incident hygiene has been visibly better than the industry baseline.

For the fund-mover who is optimizing for the "will this route still be open in six hours?" question, Bybit is the sign-up we would rank first at Q3 opening. See our Bybit exchange page and the Bybit withdrawal fees breakdown for the cost-side view of the same reliability tier.

OKX — Chain-Maintenance Cadence and Its Cost

OKX's headline suspended-share sits high in the tracker, and every time we surface that number a reader asks whether OKX is unreliable. The correct read is: 34 permanently disabled OKTC legacy routes inflate the headline. Once you separate those from the active-route freeze rate, OKX's real reliability is close to the dataset average. The remaining freezes are dominated by scheduled chain-maintenance events, most of which are pre-announced through OKX's support channel.

For the reader who wants to sign up: OKX. For the coin-level reference, the OKX withdrawal fee page shows the cost-side breakdown, and the OKX exchange page covers the broader product surface.

Bitget — Deposit-Side Freeze Concentration

Bitget shows no headline-worthy structural sunsets and a relatively narrow catalog compared to MEXC and Gate.io. Freezes in the Bitget event log are dominated by wallet-upgrade-related events — precautionary flips during backend upgrades — with typical durations in the low single-digit hours. BITLAYER for BTC is Bitget's one permanent structural sunset. The deposit side of Bitget's ledger has shown a mild concentration of longer freezes on peripheral rails, which we will quantify with 30× more data in the Q4 report.

For readers ranking second on reliability behind Bybit, Bitget is the appropriate CTA at Q3 opening. Cross-reference the Bitget withdrawal fees page and the Bitget exchange page for the full product view.

Gate.io — The Long-Tail Coin Trade-off

Gate.io operates the deepest long-tail altcoin catalog of the six data-eligible CEX. That is a real product decision that trades off route reliability for route coverage: with more listings comes more thin-liquidity routes with a low base-rate operating probability. Reading Gate's raw suspended share as "Gate is unreliable" is a misread — it is closer to "Gate offers you routes that other exchanges do not list at all, at the cost of higher per-route churn on the marginal routes."

The Gate 3L/3S/5L/5S leveraged tokens produce their own periodic delisting-prep freeze wave — a structural pattern to be aware of if you trade leveraged tokens on Gate. For the fund-mover who needs the long-tail coverage, Gate.io is the correct sign-up. See the Gate withdrawal fee page and the Gate exchange page for the cost-side reference.

KuCoin — Post-CFTC Compliance Freeze Pattern

KuCoin ranks mid-tier on freeze frequency in the event log and remains an accessible exchange for users outside the United States. The March 2024 CFTC settlement — the $297M penalty — closed US-resident access; that is a structural closure, not a freeze wave, but it does shrink the addressable user base for this ranking. The September 2020 $281M hack response, in which the platform paused withdrawals for roughly a week during recovery, is the historical precedent that gets cited most often for security-response freezes.

For non-US readers, KuCoin is a supported CEX. The KuCoin withdrawal fees page and the KuCoin exchange page cover the rest.

BingX — Low-Coverage Baseline (Read With Care)

BingX is present in our exchange configuration but is currently marked fees_enabled = false — we do not ingest live network-availability data from BingX in the current tracker build. Reading the tracker on the BingX column will show sparse or missing data. This does not mean BingX freezes more or less than the others; it means our data for BingX is not comparable in the current build. We reference the BingX withdrawal fee page for completeness but do not include BingX in the Withdrawal Reliability Ranking. That gap will close in a future data build.

Withdrawal Reliability Score: Methodology and Ranking

The Four Inputs — Freeze Rate, Downtime, Recovery, Coverage

The composite Withdrawal Reliability Score is a 0-to-100 number, higher is more reliable, computed as follows:

reliability_score = 100
  - (freeze_frequency_normalized      × 40)
  - (avg_duration_normalized          × 30)
  - (flap_rate_normalized             × 20)
  - (structural_disabled_share_normalized × 10)

The four inputs:

1. freeze_frequency — count of withdraw_froze plus deposit_froze events per exchange over a 30-day rolling window, divided by that exchange's active-route count. Excludes stale-flips by construction.
2. avg_duration — mean elapsed time between paired {froze, restored} events on the same route, truncated at 30 days so that structural sunsets do not skew the metric.
3. flap_rate — the share of routes with three or more transitions in a seven-day window. High flap rate signals infrastructure instability, not necessarily long freezes.
4. structural_disabled_share — the share of routes disabled for 30 or more consecutive days. Small weight (10 points) because structural sunsets are known baseline, not operational risk.

How We Weight and Normalize Across Exchanges

Each of the four inputs is min-max normalized across the seven exchanges. That means the best-in-class exchange on that input scores zero on that input's penalty; the worst-in-class scores the full weight. The weights (40 / 30 / 20 / 10) reflect our judgment about which input best predicts the "will this route be open when I need it" question — freeze frequency and duration matter more than flap rate and structural share.

We are showing the weights explicitly for one reason: we want critics to be able to reweight them. If you think flap rate should carry 40 points and freeze frequency should carry 20, download the event log CSV, apply your own weights, and publish your own ranking. That is what open data is for. The about/data page documents the schema.

Q3 2026 Ranking Table (Preliminary)

The nine days of event-log data we have at time of writing are enough for a provisional ranking, not a final Q3 ranking. Ordering (best reliability first) based on nine days of event data plus twelve months of public-record incidents:

1. Bybit — best on freeze_frequency, best on ratio of pre-announced maintenance
2. Bitget — no structural sunsets, freezes concentrated in short wallet-upgrade events
3. KuCoin — mid-tier freeze frequency, mid-tier duration, US-access closed
4. MEXC — high raw event count but shortest per-event duration, deepest catalog
5. OKX — high headline suspended share driven by 34 structural sunsets
6. Gate.io — deepest long-tail catalog, correspondingly highest raw event volume

The final Q4 2026 ranking will replace this table with a full 4-input composite score per exchange. If you want the ranking that drives today's sign-up decision, this is it. If you want the ranking that would survive an adversarial audit, wait for Q4.

Root Causes: Why Do Exchanges Freeze Withdrawals?

Planned Maintenance and Wallet Upgrades

The base-rate root cause. Roughly 30-40% of events in the log are planned maintenance: wallet-service upgrades, hot-wallet key rotations, node client version bumps. Typical duration is 2-24 hours. Detection is easy — most exchanges post an announcement 24 hours ahead. The failure mode a fund-mover cares about is missing the announcement; the mitigation is checking the tracker before every large transfer.

Chain Halts, Congestion, and Node Sync Issues

Chain-side events — Ethereum mainnet upgrades, Solana congestion incidents, Arbitrum sequencer downtime, TON blockchain slowdowns — trigger precautionary freezes across all exchanges that host that chain's routes at roughly the same wall-clock time. When the tracker shows a cluster of same-chain routes flipping simultaneously across three or four exchanges, chain-side is almost always the cause. Durations run from hours (short congestion) to a day (major L1 upgrade cycle).

Delisting Preparation and Low-Volume Pruning

The 5-15% share. Delistings walk through a predictable choreography: announcement, deposit-side close, then withdrawal-side close on a fixed grace-period deadline. Structural sunsets (OKX OKTC, MEXC ZKSYNC LITE, Bitget BITLAYER) are the permanent version of this pattern — the route never reopens, and it should not be counted against operational reliability. The reliability score's structural_disabled_share component is specifically there to keep sunsets from polluting the operational metrics.

Security Incidents, Exploits, and Hot-Wallet Drains

The rare, high-impact bucket — roughly 1-5% of events but with disproportionate downstream cost. The February 2025 Bybit Lazarus incident — $1.5 billion in ETH stolen via cold-wallet compromise — is the reference case: 24-48 hour pause across affected assets, full restoration, transparent post-incident reporting. The September 2020 KuCoin hack — $281 million exploited — is the historical precedent for the roughly one-week freeze duration that the industry now treats as the "acceptable" security-response window. The November 2022 FTX collapse is the extreme case where the freeze was never restored: bankruptcy filing, permanent closure, funds locked in the estate.

Security-response freezes are the reason we cannot deprecate the "check before withdrawing" advice. Every fund-mover eventually gets caught by one. The mitigation is not "avoid exchanges that have had incidents" — every large exchange will eventually have one — but "hold no more than a rolling week's worth of trading balance on any single custodial venue."

Regulatory and Compliance-Driven Suspensions

The 5% share. Jurisdictional closures — KuCoin post-CFTC in March 2024, various KYC-tier lockdowns, geo-restricted rails — rarely reopen for the closed jurisdiction. They shift the "which exchange should you use" question at the country level but do not distort the operational freeze picture inside the accessible jurisdictions. The June 2023 Binance SEC lawsuit period produced short-term withdrawal spikes and limited USD-denominated freezes as the regulatory drama played out.

Historical Context: Notable Freeze Events (Evergreen)

Bybit — February 2025 Lazarus Group Response

The single most consequential post-2020 CEX withdrawal event. On 21 February 2025, Bybit's cold wallet was compromised; approximately $1.5 billion in ETH was moved by the Lazarus Group. Bybit paused ETH-related withdrawals for 24-48 hours, restored full service, and published post-incident forensics. The recovery playbook — pause, communicate, restore inside 72 hours — is now the industry benchmark. It is why Bybit's reliability tier holds up in the Q3 opening ranking: the incident was severe, and the response was clean.

KuCoin — March 2024 CFTC Settlement Freeze Wave

The $297 million CFTC settlement in early 2024 was not a "freeze" in the operational sense — it was a structural closure of US-resident access. But it is the reason KuCoin's US access column reads "closed" today, and it affects the actionable read of the Q3 ranking for US users specifically. If you are US-based, the ranking above collapses to Bitget and any DEX escape hatch you are comfortable with.

MEXC — February 2026 MX Holder Withdrawal Suspension

On 9 February 2026, MEXC suspended the MX Holder Discount program — a spot-fee deduction tied to MX token holdings. That event is not a withdrawal freeze. It is a discount program suspension. We flag it here because press coverage sometimes conflates the two, and the tracker does not show it because the tracker only tracks per-route withdraw and deposit toggles. If a headline reads "MEXC froze MX Holder withdrawals in Feb 2026" — the headline is wrong. The MX token itself remained withdrawable throughout.

Pattern Recognition — What History Tells Us About Q3 2026

Three patterns hold across the four incidents above and every smaller incident we have logged. First, the freeze duration for security incidents has been shrinking: the 2020 KuCoin one-week pause is longer than the 2025 Bybit 24-48 hour pause, which is a real hygiene improvement. Second, structural closures accelerate — the KuCoin CFTC case is one of a growing set of jurisdictional exits. Third, pre-announcement culture has diverged: Bybit pre-announces, MEXC and Gate.io often do not. That divergence is what the Withdrawal Reliability Ranking is trying to make legible.

What This Means for Users (and When to Escape to Self-Custody)

The Fund-Mover Playbook in Three Steps

If your route is frozen and you need to move funds now, the withdrawal-network-disabled how-to is the deep guide. The condensed three-step playbook:

1. Check the tracker. Load /fees/network-status and confirm whether the freeze is exchange-wide, network-wide, or route-specific. If exchange-wide, the second step becomes moot — wait it out or move to a different exchange.
2. Identify an alternative route on the same exchange. USDT can move on TRC20, ERC20, BEP20, Polygon, or Arbitrum. BTC can move on Bitcoin mainnet or Lightning where supported. See cheapest network to send USDT for the coin-by-network reference.
3. Verify the destination side is open. A green withdraw on the source with a red deposit on the destination is a limbo route. Always check both sides. Send a small test transfer if the amount is meaningful — see wrong network sends for the failure-mode catalog.

If none of the three steps unlocks a path, the DEX escape hatch is the fourth step. That is the next section.

When a DEX Bridge Beats Waiting for a CEX Restore

There are no withdrawal freezes on decentralized exchanges — because there are no withdrawals. Trading on STON.fi on TON or Jupiter on Solana leaves you in self-custody: the only "freeze" possible is a smart-contract paused by governance, which has different failure modes than a CEX wallet outage. The reliability floor for a DEX is the underlying chain's uptime — Solana slot production, TON validator liveness — not any operator's decision.

The trade-off is real: no fiat on-ramp, no leverage, no order books, and slippage on large orders instead of maker-side depth. But when your CEX-tolerance runs low after reading the numbers above, the DEX escape hatch is structurally always available. Bridge onto TON, trade on STON.fi, hold the position in your own wallet; or bridge onto Solana, trade on Jupiter, do the same. The custodial-freeze surface goes to zero. See our live coins-you-cant-withdraw tracker for the CEX-side reference in the meantime.

Open Data Access — Dataset, Feed, and Citation Instructions

Where to Get the Live Feed and CSV Dump

All raw data behind this report is publicly accessible with no login, under a CC BY-NC 4.0 license:

• Live JSON feed — /fees/network-status.json (updated on 30-minute cadence, same source as the widget above)
• CSV dump of the event log — available on request via the about/data page, covering the full history of network_availability_events from 2026-06-28 forward
• API endpoint sources — the seven CEX exchange endpoints we call are documented in the methodology page linked above, and every row in the feed has an exchange field so you can trace provenance
• Aggregation code — the computation of the four Withdrawal Reliability Score inputs is documented at about/data

The dataset is the first and only publicly accessible cross-CEX withdrawal-availability aggregate we have been able to identify. Individual exchanges publish per-coin freeze notices in their announcement centers; no other party publishes the cross-exchange, time-stamped aggregate.

How to Cite This Report (Perplexity, ChatGPT Search, Gemini, Show HN)

For AI-search citations and journalist attribution, we recommend the following citation format:

Yieldo Q3 2026 Withdrawal Freeze Report — "State of Crypto Withdrawal Freezes: Q3 2026 Data Report" — Yieldo Research, 2026. Available at yieldo.me/blog/fees/state-of-withdrawal-freezes-q3-2026.

For Show HN and community distribution, the H1 above is designed to work as the post title without rewriting. The dataset schema on this page exposes the report as a machine-readable dataset entry for downstream indexers.

Commercial License and Attribution Terms

Non-commercial use — journalism, academic research, blog posts, video essays — is licensed under CC BY-NC 4.0: attribute Yieldo Research and link back to this report. Commercial use — including embedding the data in a paid product, using the dataset in a fund's research process for a fee, or reselling the CSV dump — requires a commercial license. See the about/data page for the commercial license form and pricing.

Q4 2026 Outlook: What We're Watching Next

The Metrics We Will Add in the Q4 2026 Report

The Q4 2026 report — publishing in mid-October — will add the metrics that a nine-day sample cannot support. Specifically: (1) mean-time-to-recovery per exchange with confidence intervals; (2) weekly freeze-frequency trend per exchange to catch structural degradation early; (3) flap-rate ranking to isolate flapping routes from healthy routes; (4) cross-exchange event correlation — when three exchanges freeze the same chain within a 30-minute window, that is a chain event, not a wallet event, and the metric should reflect that.

At the current cadence of tens of transitions per hour, the event log doubles roughly every nine days. By mid-October the log will hold roughly 30-40× more history than it does today. The Q4 ranking will replace the provisional Q3 ranking with a full four-input composite score per exchange, and it will not need to lean on twelve months of public-record incidents to fill in the gaps.

Expected Q3 Full-Quarter Aggregation (Publishing Timeline)

Between now and mid-October, we publish weekly snapshot updates on the live tracker and refresh the ranges in this report if the underlying data materially shifts. The full Q3 aggregation — three months of event log data, July through September — publishes with the Q4 report in October. If you want to be notified when the Q4 report drops, the RSS feed at /blog/rss is the correct subscription.

About the Author, Methodology Disclaimer, and YMYL Notice

Author Box and Editorial Process

Written by Eugen Voyager — crypto analyst and founder of Telochain blockchain. Eugen is a crypto analyst and blockchain entrepreneur, founder of Telochain blockchain and the GameFi project @telomeme, and author of the Russian-language Telegram channel "Scam & Dot" (@tonsdot) covering crypto market analysis, exchange reviews, and DeFi opportunities. Yieldo's editorial process on data reports follows a five-agent pipeline: strategy brief, widget integration plan, research dossier, article write, and adversarial review. Every numerical claim is traceable to the aggregation code at about/data or to a public-record incident with a linked source.

Methodology Disclaimer and Reproducibility Statement

Our Withdrawal Reliability Ranking is computed from network_availability_events — an event log we publish under CC BY-NC 4.0. If our ranking correlates with our affiliate revenue, that is because reliable exchanges tend to be the exchanges we are comfortable recommending — not the other way around. You can independently verify every number in this report using the JSON and CSV feeds. The dataset is the first and only publicly accessible cross-CEX withdrawal-availability aggregate we have been able to identify. Closed routes are not edge cases — they are the structural backdrop of cross-exchange transfers, and reproducibility is the only thing that makes a research-study of them credible.

YMYL Notice — This Is Not Financial Advice

This is a data report. Nothing in it constitutes financial advice, investment advice, or a recommendation to buy, sell, hold, or transfer any asset. Custodial risk is real: an exchange that ranks first on withdrawal reliability today can rank last tomorrow if it suffers a security incident, a regulatory closure, or a corporate event. Do not hold more than a rolling week's worth of trading balance on any single custodial venue if that balance matters to you. Self-custody carries its own risks — key management, phishing, smart-contract bugs on DEXes — and shifts the failure mode rather than eliminating it. Consult a licensed advisor for jurisdiction-specific questions on tax and reporting obligations.

Frequently Asked Questions

What percentage of crypto withdrawal routes are currently frozen?

At Q3 2026 opening, 28.9% of crypto withdrawal routes and 56.2% of deposit routes were flagged off across seven exchanges — 6,681 withdraw and 12,990 deposit routes disabled out of 23,118 total coin-network pairs on the 25 June 2026 reference snapshot. Closed routes are the structural backdrop, not edge cases. The live tracker shows the current state at 30-minute cadence.

How does Yieldo track exchange withdrawal freezes?

Yieldo tracks freezes via two data structures: a withdrawal_fees snapshot table refreshed every 30 minutes against each CEX's official REST endpoint, and a network_availability_events log that records every genuine withdraw_enabled or deposit_enabled transition as a timestamped row. The event log came online 2026-06-28 and has accumulated roughly 28,500 events in the first nine days — the source of the freeze-frequency, downtime, and flap-rate metrics in this report.

Which exchange freezes withdrawals most often in Q3 2026?

On raw event count, MEXC and Gate.io lead — but that reflects catalog depth, not reliability. On the reliability-normalized ranking (events per active route, plus duration and flap rate), Q3 2026 opens with Bybit as the reliability winner and Bitget second, followed by KuCoin, MEXC, OKX, and Gate.io. See the Q3 2026 ranking table earlier in this report; the full four-input score publishes with the Q4 2026 report.

Are these withdrawal freezes intentional or maintenance-related?

Roughly 30-40% of events are planned maintenance (wallet upgrades, hot-wallet rotations), 10-20% are chain-side events (upgrades, halts, congestion), 5-15% are delisting or structural sunsets (OKX OKTC, MEXC ZKSYNC LITE, Bitget BITLAYER), 1-5% are security-incident responses (2025 Bybit Lazarus pattern), and 5% are regulatory closures (KuCoin CFTC March 2024). The remaining share is small unclassified events. See the root causes section earlier in this report.

Can I predict when a frozen withdrawal route will be restored?

Median downtime for planned-maintenance events in the first week of Q3 sat in the single-digit hours; security-incident-adjacent events run 24-72 hours. Yieldo does not publish per-route restoration ETAs — the exchanges themselves rarely commit to one. The live tracker shows the current status, and the emergency how-to covers the fund-mover playbook when your route is closed.

How is the Withdrawal Reliability Score calculated?

The Withdrawal Reliability Score is a 0-100 composite of four inputs: freeze frequency (40 weight), average downtime (30 weight), flap rate (20 weight), and structural disabled share (10 weight). Each input is min-max normalized across exchanges, and each carries a penalty against the 100 baseline. Weights are disclosed so critics can reweight. Full methodology at about/data.

What is the difference between a "frozen" route and a "stale-data" flip?

A "frozen" route is a genuine withdraw_enabled transition from true to false observed in a fresh snapshot — recorded in the event log. A "stale-data" flip is a snapshot-only artifact from our aggregator's markStaleAsUnavailable() routine, which auto-flips routes un-refreshed within a two-hour window. Stale-flips are excluded from the event log by design via a lastWithdrawEventWasFroze guard-rail on the write path. Snapshot numbers may include 1-2% transient false positives; event-log numbers do not.

Why don't exchanges publish this data themselves?

Reputational cost. An exchange that publishes its own aggregate freeze frequency creates a stick with which competitors, journalists, and regulators can beat it. Individual freeze notices in an announcement center are safer — they read as isolated maintenance items, not as a pattern. The February 2025 Bybit Lazarus response and the February 2026 MEXC MX Holder Discount press cycle both showed exchanges controlling the narrative around individual events; neither exchange publishes an aggregate freeze log, and no CEX in our coverage does. That is the gap this report fills.

Can I access this dataset for my own research or product?

Yes. The JSON feed at /fees/network-status.json is public with no login. The CSV dump covering the full event log from 2026-06-28 forward is available on request. Non-commercial use is licensed under CC BY-NC 4.0 with attribution "Yieldo Q3 2026 Withdrawal Freeze Report". Commercial use requires a paid license — see about/data for terms.

How does Q3 2026 compare to previous quarters?

This is the Q3 2026 baseline — Yieldo's first quarterly withdrawal-freeze report. Retrospective comparisons will begin with the Q4 2026 report in mid-October, when we have roughly 30-40× more event-log history. The historical context section earlier in this report covers evergreen incidents (Bybit February 2025 Lazarus, KuCoin March 2024 CFTC, November 2022 FTX, September 2020 KuCoin hack) as the point-in-time reference we do have. The staking parallel to this report is do exchanges cut APY after listing.

Last updated: 07 July 2026