Environmentally induced errors aren't software bugs, just because there's a problem elsewhere doesn't mean we shouldn't seek to mitigate other problems.
In plane investigations I've looked at (not many) the issue has always been a compounding of several errors or shortcomings .. that strongly suggests you shouldn't let small errors build up in different systems, to me. [1]
If it's a register which takes down the whole system then surely they'd know that (and could fix it with a watchdog that returned the effected systems to the boot state without reboot) -- other comments seem to be saying "meh, it's complex, doesn't matter what the error is as long as reboot fixes it"; that seems really dangerous in safety critical systems.
[1] but I acknowledge the "better the devil you know" issue and that pragmatism and cost take over at some point.