I'm just guessing, but maybe what you're seeing is the effect of the fact that (fully) matching prefixes overwrite each other in the internal data structure of wireguard. They don't just "cover" (i.e. one is prioritized over the other) but overwrite each other fully.
This means that if another peer's entry overwrote your previous one, it's not enough to delete that peer for the overwritten one to reapoear, but the re-insertion has to be forced. The simplest is to change the allowed-addresses to something harmless then back.
I apologise, I don't understand what you're saying.
If it helps, no changes had been made in some days (at least) in the allowed-address fields, nor in any other way to the WireGuard configurations. This seems to be some persisting bug in the 7.2x ROS series' WireGuard implementations
And, what got me to look at this at all, this morning, was the fact that abruptly certain very specific traffic wasn't flowing correctly. After the reboot (with no configuration changes having been made) the traffic flowed again.
No idea, then. It might have been a connection tracking anomaly over the path - that's known to upset wireguard, and the time needed for a reboot often allows the offending entry to time out (or if it's on the router itself, it's purged.)
I don't think this has a lot to do with the enable/disable bug. RouterOS doesn't reconfigure wireguard if there's no change.