Lido & stETH Weekly News Review — Week of September 28–October 4, 2026

1. The week in one arc

This edition covers September 28 – October 4, 2026 — six dated stories, all sourced, all inside the window, and not one figure in this column drawn from a print dated after the window closed.

The week belongs to a security story. On September 30, MetaMask disclosed an infrastructure compromise inside its non-custodial staking operation and began proactively exiting the Ethereum validators it runs through Lido — roughly 17,000 validators carrying about 523,000 ETH. By October 2 the network's exit queue had spiked from ~166,000 ETH (Sep 29) to a peak above 850,000 ETH, the largest withdrawal backlog since December 2025, stretching the exit wait from about three and a half days to roughly two weeks. Around that spine: the ETF bid flipped from five green days to −$138.1M; Bitmine's September 28 SEC exhibit put its treasury above 6 million ETH; a September 29 DeFi selloff took LDO down 10.5% in a single session; and a weak October 2 jobs report collapsed the odds of another Fed hike.

ETH closed the week at $2,694.14, down 0.2% close-to-close — the flattest weekly print since this site began publishing — inside a $2,653.59–$2,715.75 range. That is the sentence that matters: the market absorbed the largest security-driven validator exit of the year without moving a half a percent in either direction, and the peg never left a 16-basis-point band.

Source: every item below carries its own source and date. Price series: DefiLlama coingecko feed, Sep 27 and Oct 4 prints read together in one session. ETF flows: Farside Investors final daily table. Network and queue figures: the dated prints named in each section.

2. Sep 30: MetaMask discloses a staking-infrastructure compromise and exits its Lido validators

At 23:35 UTC on September 30, MetaMask published a security notice: it was responding to “an ongoing security incident affecting part of our infrastructure” and, as a precaution, proactively exiting affected validators within its non-custodial staking operation — the validator business formerly known as Consensys Staking. The exits are concentrated in Lido: in a governance-forum disclosure the same day, MetaMask Staking said it had begun exiting its Ethereum validators in the protocol following an investigation into an infrastructure compromise, and that stETH holders need not take any action.

Three design details do most of the explaining. First, MetaMask stressed that it does not manage withdrawal keys for client stake — the stake is non-custodial, so the operator's exposure is to its own signing infrastructure, not to user principal. Second, a October 1 follow-up stated there was no indication that MetaMask wallets or customer funds were affected. Third, the response was sized to the worst case rather than the observed case: on-chain researcher 0xKaden's analysis, published October 1, found that block rewards from 18 of 19 examined validators had been redirected to an address funded through Tornado Cash — an attacker take of roughly 0.36 ETH, under $1,000 at print. MetaMask has not confirmed those figures, has not disclosed what was compromised, and has not said whether signing keys were exposed. Every one of those open questions is why the operator chose exit over monitoring.

The mechanism matters for anyone holding stETH. Exiting a validator does not touch the stETH in circulation: the operator's stake is withdrawn from the beacon chain, the stETH it had issued remains a claim on the protocol's remaining stake, and Lido's redemption path was unaffected all week. What changes is reward flow — a validator in the exit queue earns nothing, and its rewards stop accruing to the pool until the stake returns. That is why Lido flagged possible downtime penalties as a live risk and pointed to an ad hoc reserve fund of more than 6,750 stETH as a buffer. The cost of this incident lands on timing, not on principal.

Source: MetaMask security notice (Sep 30, 2026, 23:35 UTC) via Blockhead; Lido governance-forum disclosure (Sep 30, 2026) via Blockhead; MetaMask Oct 1 follow-up and 0xKaden analysis via Crypto Briefing (Oct 1, 2026); validator and queue figures via COIN360 and Gate research (Oct 1–2, 2026).

Take: the honest one-line summary is that a third party diverted less than a thousand dollars and the response is a nine-figure validator exit. That is not overreaction — it is what non-custodial staking is supposed to look like when an operator cannot rule out key exposure. The comparison to reach for is not a bank run; it is a flight crew grounding an aircraft after an ambiguous warning light. Expensive, correct, and decided by the party holding the least information. Watch the queue arithmetic, not the headline (the deep dive runs the numbers, and they check out to within hours).

3. Oct 2: the exit queue peaks above 850,000 ETH — the largest backlog since December 2025

The precautionary exits moved the network's queue, and the queue moved fast. The Ethereum exit queue sat near 166,000 ETH on September 29, crossed 700,000 ETH on October 1, and peaked above 850,000 ETH on October 2 — roughly $2.31 billion at the Oct 2 ETH print of $2,715.75 (DERIVED) and the largest backlog since December 2025. The estimated exit wait stretched from about 3.5 days to roughly 14 days at the peak: COIN360's Oct 2 read of 773,447 ETH in the queue carried a printed wait of 13 days 10 hours, and the churn arithmetic in this edition's deep dive reproduces that to within hours.

About 523,000 ETH of the backlog is MetaMask's exited stake across roughly 17,000 validators — 5.3% of the ETH under Lido's management (DERIVED: 523,000 against the 9.84M ETH implied by Lido's $26.51B TVL at the Oct 4 ETH print). Per the September 30 disclosure, the final affected validators were scheduled to enter the exit process by October 7, with the full exit, withdrawal and re-entry cycle estimated at up to 45 days. Lido pointed to its spread of node operators — the lido.fi scorecard's ~900 figure — and the reserve fund as the two buffers that make a single operator's exit a scheduling problem instead of a solvency one.

Source: queue path via Crypto Briefing, COIN360 and Gate research (Sep 29 – Oct 2 prints); 773,447 ETH and 13d10h wait per COIN360 (Oct 2, 2026); cycle estimate per Lido's Sep 30 governance disclosure; exit-queue value and the Lido share figure DERIVED on the DefiLlama coingecko feed Oct 4 print.

Take: a queue is not a sell, and the week proved it. Every ETH in that line belongs to an operator executing a security playbook, and the withdrawal queue processes it at the churn limit no matter how the tape feels — the backlog clears on the chain's schedule, roughly 57,600 ETH per day, which is why the printed two-week wait and the physical ceiling agree. The falsifiable marker for next week: if the queue does not fall steadily once the final validators enter the exit process, something else is in the line. Second marker: whether the returned stake re-enters through Lido or waits in the entry queue, which was 2.48M ETH on the last dated board (Ebunker, Sep 28).

4. Sep 29: a DeFi selloff takes LDO down 10.5% in a session — its second double-digit day inside a week

September 29 delivered the sharpest single-session move of the window: LDO closed at $0.43348 on our series, −10.5% on the day — its second drop of more than ten percent inside a week (Sep 24: −11.0% on the prior edition's prints). The weakness was reported across DeFi governance tokens the same session, and the two prints we hold on our own series corroborate the breadth: UNI fell 12.5% across the two sessions to Sep 29 ($9.824 on Sep 27 to $8.592).

LDO then rebounded +9.2% on September 30 to $0.47318, slid again into October 2 ($0.43952, the window's low close), and recovered into the close at $0.46354 — −4.6% on the week, on a market capitalization of $386.7M (DERIVED on the resolved 834.151M supply basis). No Lido-specific disclosure inside the window explains any leg of that path. The protocol's own numbers printed the opposite of the token's: Lido TVL ended the week +0.1% at $26.51B, the peg never left a 16bp band, and share of stake was unchanged at roughly 22.5% (DERIVED). We describe the sequence without assigning it a cause.

Source: LDO and UNI daily series, DefiLlama coingecko feed, same-session prints; LDO supply basis per Analytics Insight market data (Sep 17, 2026), resolved in edition five; TVL and peg per the same read session.

Take: two double-digit single-session drops in seven days is not a token with a news problem; it is a token repricing on thin liquidity while its fundamentals print nothing new. The fee engine did not change, the operator set did not change, and no governance action landed in the window — what did change is the ETF bid's direction and the rate path's uncertainty, the two macro legs a no-cash-flow governance claim trades on. LDO's realized volatility this week was roughly ten times ETH's while its protocol TVL moved one-tenth of a percent: that gap is the whole story, and it is a statement about holders, not about Lido.

5. Sep 28: Bitmine's SEC exhibit crosses 6 million ETH

On September 28, a Bitmine Immersion Technologies SEC exhibit listed 6,001,302 ETH in treasury holdings — the company's position crossing the six-million mark, on its stated path to a 5% target. The exhibit's companion figures matter more than the headline: 5,067,309 ETH staked, roughly 84% of the position, and a 7-day yield print of 2.62%, flat against the prior week. Against the 43,681,037 ETH recorded as staked network-wide in the same week's Ebunker print, Bitmine's staked balance alone sits behind roughly 11.6% of all staked ETH (DERIVED; both inputs dated Sep 28).

The distinction between treasury holdings and staked balance is the whole point of carrying both. Six million ETH in a corporate treasury is a call option on price; five million of it staked is a standing bid for validator slots and a standing supply of staked-ETH yield. Bitmine's yield print — 2.62% for the week — sits well above the network's own base APR on the same board (2.18%, Ebunker, Sep 28), because the company reports staking revenue on its own basis, before the protocol's fee split and before any of the timing effects that show up in a 7-day window. We print both and do not average them.

Source: Bitmine Immersion Technologies SEC exhibit (Sep 28, 2026), via MEXC research; network staked total and base APR per the Ebunker Ethereum staking weekly (Sep 28, 2026).

Take: the buyer reaching for a 5% share has added every week since the strategy launched, and this week it did so into a tape where ETH closed flat and ETF money left. Two things are true at once: Bitmine is the most rate-insensitive bid in the staked-ETH market, and its staked balance — 5,067,309 — did not move between the Sep 21 and Sep 28 prints. The accumulation is flowing into the treasury, not yet into new validator positions. When that staked number finally moves, it moves the largest single-operator share in the market, and the queue it would join is the same one MetaMask just filled.

6. Oct 2: a weak jobs report collapses the October-hike odds

The October 2 September Employment Situation report landed weak on every line: nonfarm payrolls +29,000 against roughly 90,000 expected, the unemployment rate up a tick to 4.2%, prior months revised down by a combined 60,000, and hourly wage growth at +3.0% year over year. The Fed had hiked 25bp on September 16 to a 3.75–4.00% target range, and as late as September 23–24 market pricing gave a further October move odds of roughly two-thirds to three-quarters. By the Oct 2 close, CME FedWatch put the probability of a hold at the October 28 meeting near 72–77% — a repricing the crypto tape noticed intraday, with equity index futures lifting on the print while ETH closed the week at its high.

Why a staking site runs this story: for a yield-bearing asset, the rate path is the competition. The gap between staking's user-net yield and cash rates is widest when hikes are the base case, and every probability point that migrates from “hike” to “hold” narrows the carry disadvantage without a single basis point of staking yield changing. On this week's numbers the carry table was still negative against every cash row we track — the 3-month T-bill at 3.92% (Sep 1 print), the 2-year at 4.34% (Aug 28 print) and the 10-year at 5.11% (Sep 23 print) all sit above staking's 2.21–2.24% user-net band — but the direction of travel turned for the first time since the September hike. That is a positioning change, not an income change.

Source: BLS Employment Situation, September 2026 (released Oct 2, 2026); FedWatch pricing via China Securities Journal (Oct 2, 2026); Treasury levels carried with their print dates; September Fed decision per edition five.

Take: the week's biggest macro print pushed the next move from “probably another hike” to “probably a hold,” and the ETF complex still finished at minus $138 million. Institutional flows and rate odds decoupled for the second week running: the jobs report bought the market time, not a bid. For carry-watchers the takeaway is narrower and more useful — if the October 28 meeting does hold, the next question is not whether staking yield rises, but whether the cash-rate floor falls fast enough to close a gap that has been widening since August.

7. The flows flip: five sessions of outflows worth −$138.1M

The mirror image of last week: US spot ETH ETFs posted net outflows on four of five sessions. The Farside daily table reads +$17.1M (Sep 28), −$2.8M (Sep 29), −$59.6M (Sep 30), −$55.4M (Oct 1) and −$37.4M (Oct 2) — a window net of −$138.1M (DERIVED), against +$689.8M the week before: an $827.9M swing in seven days. Fidelity's FETH led the retreat and BlackRock's ETHA was the lone steady product for most of the week. The Oct 2 session finalized at −$37.4M after provisional Friday reads had shown −$17.3M — the revision is disclosed and the weekly net uses final prints. Cumulative net inflows on the complex sit near $13.8 billion through the window's last session; for context, US spot bitcoin ETFs took in about +$82.9M over the same week.

Three sessions of that week are worth naming for what they are not. The Sep 30 outflow landed on the day MetaMask's disclosure became public; the Oct 1 outflow landed on the day the follow-up confirmed no user-funds impact; and the Oct 2 outflow landed on the day the jobs report repriced the rate path. A flow tape that keeps selling through good news and bad news alike is not trading this week's headlines — it is running a mandate.

Source: Farside Investors Ethereum ETF daily flow table (final prints through Oct 2); provisional-versus-final note per COINAlertNews (Oct 4, 2026); bitcoin ETF weekly per COINAlertNews (Oct 4, 2026).

Take: last week this column read the five green days as money parked, not gone. The honest upgrade this week is that the parking lot empties fast in both directions: a $830M swing in seven days is the same institutional foot that arrived, leaving at similar speed. What did not happen is a price consequence — ETH closed the week down 0.2%. The ETF complex is no longer the marginal price-setter it was in August, and this week is the cleanest evidence yet.

8. What the week did not move

Four things that should have moved on this news and did not, which is the most informative part of the edition. The peg. A 523,000 ETH exit through the protocol's redemption machinery is the single best stress test of stETH's liquidity, and the stETH/ETH ratio spent the entire week between +15.2bp (Sep 28) and −7.9bp (Oct 2), closing at −6.4bp — the quietest path in a month. Lido's TVL. Deposits did not flee: $26.51B, up 0.1% on the week. Share of stake. Roughly 22.5%, unchanged (DERIVED). Redemptions. No queue formed on the stETH side; the congestion was entirely on the beacon-chain side, where the protocol's churn limit — not Lido's contracts — sets the pace.

Read together, those four non-moves are the difference between a protocol incident and a protocol failure. The event was an operator's security problem; the protocol's job was to keep the peg, the deposits and the redemption path intact while a large participant left. It did all three, in public, inside a week in which LDO fell 4.6% and ETH fell 0.2%.

Source: peg, TVL and share-of-stake figures from the same single-session read of the DefiLlama coingecko feed and protocol history (Sep 27 and Oct 4 prints); queue mechanics per the Ebunker Sep 28 board and the Oct 2 queue prints cited in section 3.

9. Numbers of the week

Seven figures carry the edition; all seven are sourced and dated inside the window. ETH closed at $2,694.14 (−0.2% close-to-close on the re-read basis, the flattest weekly print in our records), inside a $2,653.59–$2,715.75 range. LDO closed at $0.46354 (−4.6%) on a market capitalization of $386.7M. Lido TVL printed $26.51B (+0.1%); the exit queue peaked above 850,000 ETH (Oct 2) on the MetaMask exits; US spot ETH ETFs finished at −$138.1M across five sessions; the stETH/ETH ratio closed at 0.99936 (−6.4bp); and the network's base staking APR stood at 2.18% on the last dated board (Ebunker, Sep 28) after a 39bp single-week step. Read together: a flat market absorbing a nine-figure security event, a flipped flow tape, and a rate path that got friendlier — none of which moved the headline price a half a percent in either direction.

Source: every figure above appears with its own source and snapshot date in sections 1–8 and the weekly data report.

10. Left out, and why

Four items a reader might expect this column to carry, and the reason each stays out. No “$2 billion exits Ethereum” panic item. The exit-queue number is real; the panic framing is not ours to launder — the stake belongs to an operator executing a security response, and the queue's mechanics are in section 3 and the deep dive. No user-funds-loss claim. MetaMask states wallets and customer funds show no indication of impact; a stronger claim in either direction is unverified, so neither gets printed. No October FOMC outcome. The committee meets October 28, outside this window; what the window carried is the jobs report that repriced the meeting's odds, and that is all this column will say about it. No post-window prints. Bitmine's next weekly update and the next network reads both land after October 4, so neither appears here; the queue's post-window path belongs to the next edition, where it will be dated on arrival. The rule is the same as the window rule: inside the dated week, or not in the edition.

Source: negative findings from the same source set; the queue mechanics are examined in this edition's deep dive.

FAQ

Related reading

• Lido & stETH Weekly Data Report — Week of September 28–October 4, 2026
• The Exit Queue Shock: What 850,000 ETH Exiting Actually Means
• stETH Yield Update — Week of September 28–October 4, 2026