The Queue Flips: 2.66M ETH Is Now Waiting 46 Days to Leave

1. The setup: from empty to overflowing in seven days

Last week's report closed with an empty exit queue: 2.074M ETH waiting 36 days to enter validation, and exactly zero waiting to leave. The reads this week, all from validatorqueue.com and all dated: Sep 10 — entry 1.895M ETH / ~33 days; exit 25,300 ETH / ~10.5 hours. Sep 12 — entry 725,000 ETH / 12d14h; exit 2,656,000 ETH / 46d3h. In seven days the network's queue machinery went from a one-way valve to a full counter-flow, the exit queue touching the level of the September 2025 panic (2.6M ETH) without the panic. This report is about what flipped, what it is worth, and what would make the flip permanent.

2. The event: one provider, one hack, one orderly exit

On September 10, Kiln — a staking provider with roughly 1.6M ETH under delegated management — announced it was exiting all of its validator nodes in an orderly fashion, following the hack of SwissBorg, to which its infrastructure was linked, in order to preserve the integrity of the assets it operates. The same statement carried the sentence everything since depends on: the ETH will "probably be re-staked with new validator keys, meaning they will not be sold." Two facts make this legible as maintenance rather than flight. First, the exodus is announced and attributed — 1.6M of the 2.656M exiting ETH has a named owner with a stated intention. Second, the entry queue's simultaneous collapse (2.07M → 725k) is what a re-stake cycle looks like from the other side of the pipe: validators leaving the active set do not need the entry queue, so new entrants slide up it faster.

Source: Kiln statement (Sep 10) via BlockBeats/WEEX (Sep 12); queue reads via validatorqueue.com.

3. What 2.656M ETH leaving looks like in dollars — carefully

At our series price of $2,513.56 (Sep 12 print), the exit queue is worth ~$6.68B (DERIVED: 2,656,000 × 2,513.56). We print our own conversion because the source's dollar figures failed the cross-check: $9.03B for the same 2.656M ETH implies $3,400/ETH, and no venue quoted anything near that inside the window — ETH traded $2,451.91–$2,522.46 on our series. The ETH counts and the wait times are corroborated across three dated reads; the dollars were not, so they are gone. The same discipline prices the eventual decision that matters: if even 10% of the exiting pile were sold rather than re-staked — 265,600 ETH — that would be ~$0.67B of sell-side flow (DERIVED), spread over the 46-day wait. Against a week in which ETFs absorbed $196.9M of net inflow in four sessions, that is a shoulder event, not a cliff. The tail risk is concentration: 100% sold is ~$6.68B, which is a different conversation.

The queue's own throughput arithmetic frames how the risk would actually materialize. 2.656M ETH over a 46-day wait implies roughly 57,700 ETH exiting per day at steady state (DERIVED: 2,656,000 / 46) — about $145M per day of queue throughput at the Sep 12 price (DERIVED: 57,700 × $2,513.56). Even the 10% sale scenario therefore does not arrive as one $0.67B block; it arrives as up to ~$14.5M a day of potential sell flow spread across the drain (DERIVED), which sits inside a single ETF session's range — this window's four sessions ran from -$29.9M to +$216.4M. Daily granularity is what turns a scary headline number into a schedulable flow.

4. The 7.9-day tail: what the wait actually costs a redeemer

The queue's 46 days is the wait to exit the active set; the delay from exit to withdrawable funds adds roughly 7.9 days (Sep 12 read) — four epochs of adjudication plus the withdrawal sweep. Total: a redeemer requesting exit on Sep 12 waits about 54 days to hold liquid ETH. This is the plumbing behind every stETH discount scare you will ever read: Lido's token trades instantly at ~$2,514.60, while the underlying redemption right behind it is a 54-day queue instrument right now. The peg printed 0.9987 (Sep 15, same-source pair) — a 13bp discount, the first in our series. That number is the market's running tally of how much instant liquidity is worth while the queue runs long. Watch it against the drain rate, not against headlines.

Aggregating the discount makes its stakes concrete. A 13bp discount across $24.29B of stETH value is roughly $31.6M (DERIVED: $24.29B × 0.0013) — what the market currently charges, in aggregate, for skipping a 54-day redemption line. For scale: about 15% of Thursday's single ETF session (DERIVED: 31.6 / 216.4) and a rounding error against the queue's ~$6.68B headline. The number is small because the mechanism is working; the watch condition is the trend — a discount that doubles week over week is a different object than one that prints once and reverts, which is why the second-print rule exists in the watchlist.

5. The concentration subplot: one balance sheet behind an eighth of the stake

While the queue churned, BitMine disclosed (Sep 8) a treasury of 5.93M ETH with 5.067M already staked. Against the 43.1M ETH now staked network-wide, that is 11.8% (DERIVED) of all staked supply — and the full treasury is ~4.9% of circulating. Ethereum's validator count keeps rising (910,898, Sep 12), which is the decentralization half of the story; the concentration half is that a growing share of those validators' delegated stake answers to fewer treasuries and fewer providers — Kiln among them, until this week. The queue flip is what provider-level risk looks like when it activates: large, orderly, and fully absorbed. The BitMine datum is what the next activation would look like if it were not orderly. Both numbers belong in the same sentence, which is why this section exists.

Converting the BitMine stack into the network's own units: 5.067M staked ETH, if sliced into standard 32-ETH validators, corresponds to roughly 158,000 validators' worth of stake (DERIVED: 5,067,000 / 32) — against 910,898 active validators, on the order of one in six slots answering to a single treasury's delegation (DERIVED; order-of-magnitude by construction, since treasury stake need not be sliced into 32s). The qualifier matters: delegation through operators spreads the keys across many machines, so operational decentralization and beneficiary concentration are different axes. The queue flip demonstrated that the operational axis works; the beneficiary axis is what a non-orderly event would test.

Source: BitMine disclosure (Sep 8) via Coinliva/Fixygen; validatorqueue (Sep 12).

6. The whale and the machine: 10,000 ETH through the same pipe

The same week, a tracked address redeemed 10,000 ETH from Lido in two 5,000 ETH instalments (Aug 26 and Sep 8), still holding 238,000 stETH — about 4% of that position redeemed over three weeks (DERIVED), with part of the proceeds re-deposited into Morpho. The scale is three orders of magnitude below the Kiln churn; the mechanics are identical. Every redemption — whale or retail — waits in the same 46-day line, and the queue does not care who is standing in it. That indifference is the design working. What the Sep 12 counter-read (exit zero, entry 1.858M) would have shown, had we carried it uncritically, is how easy it is to describe the opposite machine; we print the conflict and carry the read the named event corroborates.

Source: Ai Yi (@ai_9684xtpa) via GetChainNews/ChainCatcher (Sep 8); counter-read via CryptoNewsToday (Sep 12).

7. Three scenarios for the 46-day wait

The scenarios share one property that makes them useful: each is falsifiable with a dated public read within two weeks. No narrative required — the queue prints its own verdict.

8. What we watch (the falsifiable list)

• Exit-queue drain rate — below ~100k ETH by ~Sep 26 confirms scenario A.
• Entry-queue refill — above ~1.5M ETH re-tightens the entry wait past 30 days and re-arms the admission-lag arithmetic of last week's report (1,800 validators/day at the old flow; ~57,600 ETH/day if the entry queue refills at its prior rate).
• The peg pair — a second same-source print at or below 0.9987 upgrades the discount from data point to trend; a print back at 1.0000+ closes the question.
• Kiln's own disclosure — the re-stake claim is a statement of intent from an interested party; its confirmation is evidence, its absence is a finding.

9. Conclusion: the valve works, and that is the story

For three weeks this site reported an exit queue at zero as a conviction signal. This week supplied the counterweight: the queue can go from empty to 2.66M ETH in seven days without a single ETH being sold — because a provider asked it to, because a client got hacked, because the machinery exists precisely so that a 1.6M ETH operational problem becomes a scheduling problem instead of a solvency one. The bearish reading of the week (46 days to leave!) and the bullish reading (nobody has to sell to leave!) are the same datapoint viewed from either end of the pipe. The number that settles it is not the queue's length but its drain rate — and that prints daily, for free, to anyone watching.

FAQ

Related reading

• The Empty Exit Queue: 2.07M ETH Is Waiting 36 Days to Earn Less Than Cash
• Lido & stETH Weekly Data Report — Week of September 7–13, 2026