It seems to happen once you either trigger a node that uses the database credentials or test the connection from credentials, it will succeed the first time and then it will through the error message if you test the connection again. For the workflow, this is just a test workflow we have, the input of the SELECT node is the manual trigger, not really consuming anything from the previous node. Let me know if that answers your question.
This has been fixed in the PR below (or at least it is all good with my testing), This has been reviewed and will be merged and released at some point this week.