As far as I can tell, any datasets that use the same source and credentials will literally use the same recorded credentials, such that if something goes wrong with those credentials (such as you getting locked out of the account for some reason and the credentials being denied because of it, or if your database was briefly offline or otherwise inaccessible during a recent attempted refresh), it goes wrong for all datasets that use them. They will all show the same error even if they are not set to auto-refresh because the credentials are still used by that inactive connection. The one dataset on which you have turned off auto-refresh still references the same credentials even if it is not actively using them to access data, because it did use them to access the data in the first place and will need to if you ever manually refresh it, so it is merely mirroring the error message it shares with every other dataset that references those credentials.
The good news is that in my experience you can also fix them for all datasets at once. If you re-enter them and re-enable refresh on one set, the rest will pick up the fixed credentials when you move to re-enable their refresh. There is usually a brief delay though when you view each dataset's refresh settings, where it may still momentarily claim that there is a problem with the credentials before it catches up and that warning goes away. So you should be able to re-enable refresh, or to manually refresh everything after entering the credentials only once on one of the disabled datasets.
