Correct, sir! And mine would not be working the way they (and yours) currently are if I hadn’t kept them at backlevel firmware.
I’d be curious to know your results if you ever have opportunity to put two identical cams side by side with one at the current firmware level and the other at v4.10.3.40. I would be willing to perform such a test if Wyze saw fit to loan me a known working lab cam.
Here’s where cloud standards were first applied to local recording… Never seen anything to the contrary of it either:
My sense has always been that this was implemented to either support Person Detection accuracy, incentivise folks towards CMC, or simply to be consistent. I’d love to feel confident climbing aboard the current firmware train, but past testing + recent bad press = very reluctant me to “fix” what ain’t broken… 
EDIT-1: Based on what you’ve described, it almost sounds as if local recording works OK if you START OUT on later firmware, but breaks for some of us who had been using firmware from before the change. I know that most cams ship with MUCH OLDER firmware oob, but if you immediately upgrade it BEYOND .40 before using it, perhaps local recording starts out on better footing… 