you've suggested going further than this #4935 (comment) - but I think that's a potential follow-up. This quick-fix will be a significant improvement as-is - OK?

I don't think #4935 is much of an issue (at least not much of an 8.0.0 issue) so happy with whatever.

(the "broken platform" issue mostly impacts developers and if clean fails it makes little difference to uses how many platforms it cleaned before bailing, either way they are going to have to re-run the command later).

LGTM; just needs a test.

I don't think #4935 is much of an issue (at least not much of an 8.0.0 issue) so happy with whatever.

I agree it's a small issue. #4949 is more important, and this sort of caused that (i.e., a remote clean failure in one workflow prevented cleaning of a bunch of others at the same time). However, it still a bug (cleaning a valid remote should not depend on whether or not we try to clean an invalid one first) and easy to fix.

However, as this one is not very important and it's not instantly clear to me how to test it, I'll bump it back to 8.x

Been draft for a while!

In the mean time this functionality has been implemented as part of other work. Remote cleans are now performed in parallel with any errors reported at the end, so one failed remote clean no longer blocks others.