Unexpected failures on windows runners

Initially noticed on https://github.com/bazelbuild/rules_rust/pull/879#issuecomment-894896993

Did something change about the windows runners recently? As you can see in that comment, build [4029](https://buildkite.com/bazel/rules-rust-rustlang/builds/4029) succeeded but build [4046](https://buildkite.com/bazel/rules-rust-rustlang/builds/4046#73d9f1e8-780c-4833-bb42-85b1c29c1fa8) failed when no changes had been made to that branch.

Start with the linked pull-request comment and compare Buildkite builds 4029 and 4046, focusing on the Windows runner details shown there. Trace what changed between the successful and failed runs; done means identifying the runner-related cause and documenting or restoring the expected CI behavior.

Written by the indexing model from the issue text.