This was not a test of solvency. It was a test of token delivery.
The three books that make that distinction visible are $BNB, the venue token, $BTC, the deepest reserve line, and $USDT, the stablecoin an issuer can freeze. Withdrawals resumed at 06:00 UTC on November 1; in the first nine and a half hours the venue processed 4.649 billion dollars of withdrawals against 1.01 billion dollars of deposits, a net 3.64 billion dollars (Binance withdrawal update, November 1), about 2.1 percent of the 171 billion dollars in its October 28 proof of reserves. DefiLlama later showed 7.03 billion dollars of net outflows over 24 hours, while the next nine largest exchanges by tracked reserves took in roughly 750 million dollars combined (DefiLlama via Odaily, November 2; MarketCapOf, November 2). My own calculation: about 11 percent of that outflow appeared as net inflows at rival venues, so roughly 89 percent never rotated to another exchange ledger. On Square's November 6 trends page, withdrawals reopening drew 4,493 discussants against 1,793 for 98 percent replenishment, 2.5 times the attention (Binance Square Trending Topics). The crowd was pricing a queue, not a balance sheet.
The mechanism explains why the two questions differ. Replenishment is not recovery. Binance said it would absorb the loss from corporate funds, SAFU and its futures insurance fund, and that by November 4 about 98 percent of affected assets had been bought back through market and over-the-counter channels; the remaining 2 percent is low-liquidity tokens that take longer to source (Binance recovery update 7, November 4). Separately, Tether froze more than 1.567 billion USDT and Circle about 823 million USDC within an hour of the outflows (Crypto Briefing, October 29). A freeze stops a token moving; it does not return it. $BTC has no issuer switch, $USDT does, so the frozen slice and the replenished slice sit on different ledgers. Buying replacements in thin markets is not a balance-sheet entry: it is a purchase executed in an order book, at a price, over time. Ninety-eight percent and one hundred percent are not the same operational state.
The reserve proof has the same shape. The November 4 snapshot expanded coverage from 10 tokens to 44, 4.4 times the earlier list, and its zero-knowledge proof went live on November 6 (Binance recovery update 7; OneBullEx; CoinCamps, November 6). A Merkle tree plus a zero-knowledge proof establishes inclusion and reserve sufficiency for covered balances at one instant. It does not enumerate off-chain obligations, derivatives positions or pending legal claims, and it does not model every holder withdrawing at once (OneBullEx, November 6). A snapshot is not an audit.
The phased reopening is the case. On November 1 all 15 first-phase networks were live by 12:36 UTC, but only specified assets on each; on November 3 a further tranche covering BSC, ETH, SOL, XRP, SUI, Arbitrum, Optimism, TRX and Base came back at 12:00, with the announcement repeating that not every asset opens at once (Foresight News incident timeline, November 3; Binance Announcement). Coverage then went from 10 to 44 tokens, and withdrawals were reported restored for 62 tokens by November 6 (The Digital Track, November 6). The headline figure moved from 90 to 95 to 98 percent across those updates (Binance, October 31 to November 4). Every step changed what one holder could do; money and deliverability moved on different clocks all week.
What most people get wrong is reading 98 percent as a statement about people. It is a statement about value. Two percent of a roughly 6.3 billion dollars loss is about 126 million dollars, and it sits in the least liquid tokens, precisely the assets whose holders cannot sell into a thin book while they wait (mabonchain, November 4). The residual risk is concentrated, not evenly spread. The percentages also do not line up: the October 31 update said 90 percent, the November 2 update said 95 percent, and the November 3 incident report described 98 percent as of October 31 (Techish, cited by mabonchain). Those cannot all measure the same thing on the same date. None of this shows balances are missing; the timeline is the venue's own accounting, not an independently reviewed result.
My read is waiting, and it is refutable. I am not treating this as insolvency, and I am not treating 98 percent as an all-clear. If the 44-token proof holds, the last 2 percent is sourced without a dislocation in those thin books, and 24-hour net flows turn positive, I would drop this view and treat it as a clean operational recovery. If net outflows re-accelerate now that the proof is live, I flip: the market would still be pricing solvency, and my framing would be wrong.
Three things I will check next week, each on a public page:
1. Reserve ratios, not the headline. On the proof of reserves page, write down which of your tokens are among the 44 and the ratio shown for each. Watch the next recovery update for whether the final 2 percent closes and full withdrawals return for those assets. If it closes without a price dislocation in thin tokens, the delivery constraint is gone.
2. Flow direction, not the cumulative total. Read the DefiLlama 24-hour net flow against the 3.64 billion dollars of the first nine and a half hours. A decaying outflow is a queue clearing; a second wave is a different story. The 11 percent figure says most of that money never landed on another exchange ledger.
3. The enabled asset list, not the anniversary. Count how many tokens each withdrawal update enables and compare spot turnover before and after. If a token is enabled and its book stays thin, the queue has moved into the order book rather than disappeared.
Data as of today. Not financial advice.
Next week, which clears first: the last two percent of tokens, or the withdrawal queue for the people who hold them?
$BNB $BTC $USDT #BinanceSquare #BinanceReopensWithdrawalsForMoreAssets #Binance98PercentReplenished