lint-nix action is currently unreliable

[Renegade334]

Over the last couple of days, lint-nix has been recurrently silently failing. Unlike a normal workflow run which returns green after ~30sec, failed runs don't seem to get anywhere and are being euthanised after the 15min GHA timeout. These count as CI failures, but don't have a red X in the actions viewer, which makes them quite annoying to track down.

Looking through the output when running with --verbose, it seems like the formatter is just hanging partway through traversing the tree for some reason. Switching the runner to ubuntu-latest appears to resolve it, although I don't know how robust this "fix" is.

cc @nodejs/actions

[aduh95]

Switching the runner to ubuntu-latest appears to resolve it, although I don't know how robust this "fix" is.

Probably not much, ubuntu-latest does not have the 15-minute timeout, but that's hardly a solution, more like a workaround. A "better" workaround would be to split the linter job so we only run lint-nix on PRs that touch the .nix files, so the problem only affects Nix-related PRs.

It seems to be timing out for about 9% of the times (probably underestimated, as I just browsed the action list and didn't check for retries), and only since 5b5f069, so it looks like a regression.

[Renegade334]

FWIW, the 15 minutes is somewhat arbitrary – from the timestamps, the timing-out runs aren't going slowly, they're traversing the tree at normal speed until they suddenly hang, and the vast majority of the run is just spent waiting for the process to be killed.