I do not know how to say this politely, so I ask your forgiveness in advance.
A poster hangs over my workbench with a list of Debugging Rules, the third one of which is:
Stop thinking and look
Without knowing the actual failure, we can spend the next month batting ideas back and forth, but that will not fix your machine.
You now have a test case triggering the problem, you have a list of places where the problem may lurk, so it’s time for you to get in there and learn something new. When you track down the loose or broken part, whatever it may be, the reason it causes the effects you’re seeing will become obvious.
So: stop thinking and look.
The reason I have that poster is because all too often I start trying to reason about a problem before collecting enough evidence to support any conclusions. It’s a hard lesson to learn! 