research/experiments/002-buffer-retention/results/2026-09-22-linux-aarch64/audit.json


1
2
3
4
5
6

{
  "status": "superseded_by_corrected_measurement",
  "reason": "All rusage RSS samples were identical at 28753920 bytes. Linux resource usage survives exec; this measure includes pre-exec history and is unsuitable for child image RSS comparison. The corrected run records proc/self/status VmHWM as a separate approximate metric and retains raw rusage. A 64-slot sparse case is added to examine the latency/queue confound. Other first-run observations remain exploratory.",
  "next_run": "2026-09-22-linux-aarch64-v2",
  "sources_preserved": true
}