Presently, we have validation of readConcern that occurs in a few places in the code. This can lead to some unexpected consequences, such as read concern validation occurring later than expected (see SERVER-47444).
In some cases, this separation of validation seems unavoidable, e.g., the service entry point level validation versus the strategy.cpp validation, or the validation that must occur on transaction unstash.

We should investigate whether it's possible to merge the validation that occurs at the service entry point level with the validation that occurs in the replication coordinator, provided that the read concern is not expected to change between these two code locations.