390×844, independently reproduced in Chromium and WebKit, production dist.

Intercept the two Squeekchain endpoints with fresh generatedAt/observedAt timestamps and installationId browser-fixture; status has measurement.freshness online, rpmMilli 0, and builder.status available.

Return status available, recentSettlements [], mining.recentCheckpoints [] and mining.nextRound.round 101 from settlements.

Enter the synthetic token 123456789:ReviewOnlyFakeToken000000000000000 and chat ID 7734128; leave New round settled enabled, mock Telegram sendMessage as HTTP 200 {ok:true}, and click Start watching.

On the next 15-second poll return one safe settlement with previousSettledThroughRound 100, settledThroughRound 101, workAdded 42 and a valid blockTimestamp.

Expected: one Round #101 settled message and timestamped log row.

Actual measured: 0 sendMessage requests, empty log, while history has advanced to #101.

Return round #102 on the following poll: exactly one request and a Round #102 settled row appear, confirming that #101 alone was silently discarded.

390×844, independently reproduced in Chromium and WebKit, production dist, fresh mocked API responses every 15 seconds, installationId browser-fixture, Telegram sendMessage mocked as HTTP 200 {ok:true}.

Enter synthetic token 123456789:ReviewOnlyFakeToken000000000000000 and chat ID 7734128; turn the RPM threshold trigger off and keep round history unchanged.

Start watching with 100 RPM, return 0 RPM on the next poll and keep returning zero.

Turn Hamster sleep off, allow a full poll, then turn it on again.

Advance another 22 polls (330 seconds) with fresh timestamps and zero RPM.

Expected: one Hamster sleep alert after five minutes following re-enable.

Actual measured: sleep switch checked, 0 Telegram requests and 0 log entries after more than five minutes.

Independently reproduced by starting with sleep off, observing 100→0, enabling sleep, and advancing 22 more polls: again 0 requests.