Latest context: https://digitalasset.atlassian.net/browse/DA-2467

The problem with this flow is that it is a bit brittle. We move the party, then attempt to import the ACS. If we fail now and restart, on the next run we see that the party was migrated and don’t attempt to re-import the ACS again.

The best idea about this I have: Check the size of the ACS on the participant. If it's empty, at least log a warning, at least if it's the validator party. Perhaps also just try resume the reonboard from key in such cases (perhaps make this controllable via a config flag).