Is there somewhere you've read this? I have not heard about reboots having anything to do with reducing the risk of kernal panics. I don't know that you can accumulate flipped bits that eventually just get to be too many, or flip a bit that doesn't manifest a panic until some time in the future, but if you just reboot soon enough you'll avoid the problem.
A bit flip is either going to cause a panic, cause a data error, or be in a place that won't do anything of consequence. I feel like starting here https://en.wikipedia.org/wiki/Dynamic_random-access_memory#Error_detection_and_correction and then falling down the rabbit hole of reading actual research is probably the only way to get real data rather than opinions.