CI Flake for Windows?

# 🐞 bug report

### Affected Rule

N/A

### Is this a regression?

Unsure

### Description

Windows CI seems to be a bit flaky (flakey? what's the preferred spelling? Well added both here for future searches lol).

In #3179 I had @jbedard push a couple empty commits. In order:

1. **Passing** CI for the last real commit in the PR https://github.com/bazel-contrib/rules_python/pull/3179/commits/8476820c710f122161577720c18e2b8e8ef4e144: https://buildkite.com/bazel/rules-python-python/builds/12586/steps/canvas
2. Pushed empty commit resulted in CI **failure** https://github.com/bazel-contrib/rules_python/pull/3179/commits/0910362b4bdaf32422b21b6fe9d6d65c7dc2bd06: https://buildkite.com/bazel/rules-python-python/builds/12587/steps/canvas
3. Pushed another empty commit resulted in CI **pass** https://github.com/bazel-contrib/rules_python/pull/3179/commits/fdbb53eadee1af1fa0910836db54dac2c76857e0: https://buildkite.com/bazel/rules-python-python/builds/12588/steps/canvas

Start by comparing the linked Buildkite runs 12586, 12587, and 12588 for PR #3179, focusing on what changed between the passing and failing empty commits. The issue does not identify a failing step or expected fix, so completion criteria must be established after the intermittent Windows CI cause is isolated.

Written by the indexing model from the issue text.