Yes, this is what I thought was the strangest behaviour too. Once I have stepped through the various errors and it becomes stable, it remains stable. Even if I restart the computer. It is only when I delete the database and open up a new copy that it throws the errors again.This behavior bothers me the most. It ALMOST seems to indicate that you must have "side effects" in your code that initialize things as you go, and that because of error handling, the Runtime version doesn't catch the errors correctly. In this context, I use "side effects" to mean that something attempts to update a global variable after the code has started, and something about the error handling or external references is handled differently between Runtime vs. full Access.
I'm not going to swear this is your problem, but that description makes me think it is happening that way.
Do you have something that late-binds some definitions or something like that?
EDIT: Second question: Does this "stepwise" improvement occur as you make these tries in the same session? If so, do you get that stepwise improvement if you log out and log back in between failures?
I will look through my code and check how the global variables are handled (there are only two) and also any late binding. Would there be a reason that these would be OK when the database is just one part, but then become troublesome once it is split? That also seems really strange to me.