I can say for sure though that Defender at times has a noticeable performance impact. That's why years ago I went looking for performance comparisons in the first place.
Thanks for the informative response!
So, why are we doing scan-on-read (with substantial performance overhead) when we should instead be doing scan-on-write (when scanning can, in most cases, be done in idle CPU cycles)?
2) removable storage devices
3) the system drive is not controlled during reboots
You could imagine building a system that tracks which files we wrote and with which virus database version, which resets things to be scanned across reboots and virus database updates, and has exceptions for removable devices and so on, but it screams "attack surface"...
I always felt the same way about daily/weekly scans. How would anything get there if your client, server, etc all have AV? At that point it probably wouldn't be caught anyway.