The macOS runner image now NTP-steps the guest clock synchronously before jobs run, so GitHub step timestamps and resource-metric samples end up on the same timeline. Previously a freshly-booted VM could be handed a job seconds after boot, before macOS timed had corrected a clock that started minutes behind real time, so step start/end times were stamped minutes early relative to the metrics and only drifted back into alignment as timed converged mid-run. The image now runs sudo sntp -sS -t 5 time.apple.com right after dispatch succeeds (network is already up at that point) and before the GitHub runner and metrics sampler start, so the step timeline and the charted metrics share one timeline from job zero instead of being out of phase for the first minutes.
Hive
MacOS runner clock is NTP-stepped before jobs start, so step times and metrics align
Published
Jul 01, 2026 · 14:47 UTC
Repository
tuist/tuist