Just want to point out that breaking video output due to ABI with an OS point release upgrade is just as bad as breaking a filesystem driver. Computing is complex, and complexity is hard. No amount of OS zealotry gets around that.

No, no! It is the problems of hardware manufacturers and Linux DRM guys.
If hardware vendors provides drivers (OSS and/or proprietary) for FreeBSD, the KBI would be much, much more stable within single major release.
But as non-nvidia GPU manufacturer does not provide direct supports for FreeBSD (but thankfully re-licenses minimum required codes as-is under BSD-compatible license), FreeBSD project needs to implement compatibility shim for Linux (as LinuxKPI), which is quite moving goal.
See proprietary GPU driver by nvidia (excluding DRM driver).

I follow the root cause. A similar thing happened with RHEL and openzfs breakage due to KBI change in a point upgrade recently. When that happened the zealots said such a thing wouldn’t happen on BSD. Yet it did. In both cases it’s surprising that such a breakage wasn’t caught prior to release. Complex systems are hard and no system is immune to these complex interactions.