Spurious CI failures on Darwin due to nightly reboots
Currently the Darwin runners reboot on a nightly basis to clear Rosetta's JIT cache, which otherwise grows without bound.
However, this has the unfortunate effect of killing whatever job is unlucky enough to be running at the time. We should either:
- find another way to clear this cache
- ensure that the reboot waits until the currently job has finished
- kill the job but ensure that it is restarted on another runner