Currently, the release process involves running releasebot first in prepare mode, then again in release mode.

For release types other than beta, a CL that updates the VERSION file is created in prepare mode. It is also tested (soon, including longtest builders, see #29252). After that CL is submitted to the release branch, the release mode takes the latest commit on the release branch and produces a release, skipping all testing (because thorough testing was done in prepare mode).

There is a possibility of a change being submitted to the release branch between the time the CL updating the VERSION file is created and tested in prepare mode, and the release mode picking it up. That would mean the CL that landed would bypass all testing done in prepare mode.

This issue is low priority for two reasons:

1. The chance of this happening is low because only a small set of release managers have access to submit changes to release branches, and we are generally aware when a release is happening and manually avoid touching the release branch.

2. Additionally, changes to release branches undergo testing (soon, including longtest builders, see x/build/cmd/coordinator: include -longtest and -race builders for trybots on release branches #37827), so there isn't a risk of introducing a problem that would be caught by testing.

However, it would still be unintentional if it happened, and it might require release notes to be updated. This is the tracking issue to investigate and implement a way to remove this small possibility of a human mistake. It was split off from #36182.

/cc @toothrot @cagedmantis @andybons