Shards retry ShardCannotRefreshDueToLocksHeld locally one single time after refreshing the catalog cache, but if on the retry it hits the same error (because the collection metadata was invalidated again after the refresh) then the ShardCannotRefreshDueToLocksHeld error is propagated to the router.

The router should retry on this error, but it doesn't because it isn't in this list of errors to be retried on write_op