Wait long enough before calling a boot failed
scope: generic · severity: trap · confidence: proven · subsystem: boot
On taimen, sshd does not listen until about 111 s after power-on. Verdicts were called wrongly at 75–90 s more than once — each one a working kernel recorded as a failure.
Wait at least 180 s before calling a boot failed, and measure your own
device’s number rather than inheriting this one. Record it as
PORTHOLE_REBOOT_BUDGET_S in the profile.
The number is not the point. The point is that “how long until userspace answers” is a measurement you must take once, and until you have taken it every boot verdict you issue is a guess. Take it on a known-good kernel, on a cold boot, and write it down.
Note this is a different quantity from a warm reboot cycle, which on taimen was ~30 s forced and ~45 s graceful.
Related: poll-never-sleep, prove-which-kernel-answered.
