Point in time recovery: Recovery target is unreachable in combination with logging_collector = off

PostgreSQL 13 has changed the behavior when the recovery target is unreachable while cloning clusters. A fix has been implemented such that the behavior remains as in previous versions (see Spilo pull request)

However, this fix relies on screening the log files - something that doesn't work when logging_collector is set to off (e.g. because logs shall be collected by Promtail).

My current workaround: Set logging_collector to on during cloning and switch back back to off later

Since this cost me half a day, this issue serves two purposes :)

1. maybe someone else has the same issue and it might be easier to find the solution in an issue in the postgres-operator repository instead of a comment in a pull request of the spilo repository
2. I would suggest to add a brief note in the documentation of the postgres-operator (e.g. in https://postgres-operator.readthedocs.io/en/latest/user/#how-to-clone-an-existing-postgresql-cluster)

Start with the cloning guidance at user/#how-to-clone-an-existing-postgresql-cluster and review the linked Spilo pull request for the recovery-target behavior. Document the limitation when logging_collector is off, including the workaround described here; done means users can find and understand this caveat without searching the Spilo issue thread.