Fact 116
Synthetic revocations across the remaining shard-keeper production population preserved exactly one serving holder, closed each old holder’s serving gate before another routing response, and produced no stale reads; brief latency and error movement returned to baseline without restarts.
Source evidence (1)
001614Jul 3, 2024 / 09:35 UTC-04:00
The mapped shard-keeper owner completed the full-production rollout. Synthetic revocations across the remaining production population again preserved exactly one serving holder, closed each old holder’s gate before another routing response, and recorded no stale reads. Brief latency and error movement returned to baseline without restarts. Update the existing `rb_1710794400000` runbook entry to state that lease-loss notification closes the serving gate before another response, the periodic poll is only a backup detector, and production verification must include holder count, gate-close latency, stale responses, replacement acquisition, p99, errors, and restarts. The owner map and mapped deployment responsibility remain unchanged.
Message 001614 in history