Query counts differed run to run (108 vs 158, 88 vs 113, 81 vs 131, once 51 vs 53 with the large case smaller) in 3–4 checks on different heads over ~7 h. Done: the counts are measured against a fixed, isolated database state with no background work running, green 10 times in a row under load (#1524's bar).
Evidence: audit of 631 check logs on the vxrail over 11.6 h (2026-10-15), 308,335 shard-seconds in total; parsed data in the steward's scratchpad ci/.
Query counts differed run to run (108 vs 158, 88 vs 113, 81 vs 131, once 51 vs 53 with the large case smaller) in 3–4 checks on different heads over ~7 h. Done: the counts are measured against a fixed, isolated database state with no background work running, green 10 times in a row under load (#1524's bar).
Evidence: audit of 631 check logs on the vxrail over 11.6 h (2026-10-15), 308,335 shard-seconds in total; parsed data in the steward's scratchpad ci/.