cylc clean: continue if remote clean fails?

[hjoliver]

If cylc clean can't clean a remote job platform used (or attempted) by the target workflow, it should continue on and clean other job platforms, and locally, rather than abort immediately.

I'm reasonably sure this is a bug and not a feature(?)

E.g. install and run a workflow that uses a non-existent job platform, then try to clean it.

[hjoliver]

see also #4673

[oliver-sanders]

It should continue onto other platforms, yes, however, it should not clean locally because doing so would delete the database prohibiting cylc clean from being able to remove the remote directory at a later date e.g. if a login node goes down.

[oliver-sanders]

If you try to run a task on a non-existant platform, then, ideally, cylc clean should not attempt to clean that platform since it never installed on it in the first place. But cylc clean does not have access to the remote-init map, perhaps it should, it wouldn't hurt to store remote-init events in the DB?

[oliver-sanders]

(taking a quick look at the broken platform side of this)

[oliver-sanders]

This has now been fixed by other changes, see #4950 (comment)