Have you looked at the communication on a scope to see if there is a difference?

Or have you augmented your code and had the code measure timing differences etc?

Dummy-staring at code is best with everything is stone-dead, or when you have already used a process of elimination and figured out that it is a specific routine that does something wrong for some specific stimuli.

In your case, you have different function with two different test cases. So your job is to figure out what differs. Such analysis is part of developing code (or hardware). Don't ask on a forum without first spending some time yourself. We can't measure on your hardware. We can't run test code on your setup. You can. Take advantage of that.