Tracker for CI failures
This issue is intended to serve as a reference for known sources of instability in the CI. Jobs should not spuriously fail, and you should not need to retry in order to have a passing pipeline.
If you see failures which seem spurious (i.e. not caused by your code), please read the comments below to see if anything is related, and add a link to the failing job as well as a summary (e.g. 'machine didn't boot', '$driver GLES test failed but I only changed Vulkan code', etc) so we can track and fix test failures. Thanks.