GitHub's default timeout is 6 hours. Recently some of my GitHub Actions
jobs have started randomly stalling for that long, which is inconvenient
because it ties up a chunk of my runner quota. It apepars to be very
rare for a job to recover after stalling. It's better to time out
quicker and retry on a different runner.
The old version available by default in the GitHub Actions image fails
with:
thread 'main' panicked at 'package cache lock is not currently held, Cargo forgot to call `acquire_package_cache_lock` before we got to this stack frame', /usr/share/rust/.cargo/registry/src/github.com-1ecc6299db9ec823/cargo-0.57.0/src/cargo/util/config/mod.rs:1542:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
For cron scheduled runs, GitHub Actions names the run after the workflow
name (unlike Travis which names it after the subject of the most recent
commit on the branch). Having a bunch of runs called "test" made it look
like someone was poking at things manually to debug something, rather
than these being ordinary cron runs.
The crontab "40 1 * * *" is once daily at 01:40 UTC, which is 100
minutes after nightly promotions are kicked off so hopefully the newest
nightly is available by then.
