We've seen a few failed tryjobs for our win_compile_rel trybots (AFAIK it only hits these, not sure why):

https://build.chromium.org/p/tryserver.webrtc/builders/win_compile_dbg/builds/7316 (vm207-m4)
https://build.chromium.org/p/tryserver.webrtc/builders/win_compile_dbg/builds/7319 (vm495-m4)
https://build.chromium.org/p/tryserver.webrtc/builders/win_compile_dbg/builds/7320 (vm207-m4)
https://build.chromium.org/p/tryserver.webrtc/builders/win_compile_rel/builds/389 (vm363-m4)
https://build.chromium.org/p/tryserver.webrtc/builders/win_compile_rel/builds/392 (vm263-m4)
https://build.chromium.org/p/tryserver.webrtc/builders/win_compile_rel/builds/397 (vm212-m4)

The errors look like this:

===Running git.bat remote set-url origin E:\\b\\c\\git_cache\chromium.googlesource.com-external-webrtc (attempt #1)===
In directory: E:\b\c\b\win_compile_rel\src
fatal: Not a git repository (or any of the parent directories): .git
===Failed in 0.0 mins===

...

Something failed: git.bat remote set-url origin E:\\b\\c\\git_cache\chromium.googlesource.com-external-webrtc failed with code 128 in E:\b\c\b\win_compile_rel\src after 5 attempts..
Ran 327.322000027 seconds past deadline. Aborting.
    git_ref = git_checkout(solutions, revisions, shallow, refs, git_cache_dir)
  File "E:\b\rr\tmprlt0sp\rw\checkout\scripts\slave\.recipe_deps\depot_tools\recipe_modules\bot_update\resources\bot_update.py", line 502, in git_checkout
    git('remote', 'set-url', 'origin', mirror_dir, cwd=sln_dir)
  File "E:\b\rr\tmprlt0sp\rw\checkout\scripts\slave\.recipe_deps\depot_tools\recipe_modules\bot_update\resources\bot_update.py", line 228, in git
    return call(*cmd, **kwargs)
  File "E:\b\rr\tmprlt0sp\rw\checkout\scripts\slave\.recipe_deps\depot_tools\recipe_modules\bot_update\resources\bot_update.py", line 213, in call
    code, outval)
__main__.SubprocessFailed: git.bat remote set-url origin E:\\b\\c\\git_cache\chromium.googlesource.com-external-webrtc failed with code 128 in E:\b\c\b\win_compile_rel\src after 5 attempts.
step returned non-zero exit code: 1


We haven't done any recent changes infrastructure-wise that could explain this...
Since it's so many different machines, I don't think it's just a faulty disk/unlucky corruptness.

Is there anything that can be done to easily wipe these caches for all the slaves?