ReFS doesn't give you any more rollback capability than NTFS in that sense. ReFS supports file level snapshots, not volume.
And on a client machine, it's of much less importance overall (to you, it may be super important and I don't want to discount that). And on the server side, that's why we have n+2 failovers. No single machine of importance should ever be a point of failure... I realize that's not always reality but it's more or less Microsoft's position; after all, why sell one Windows Server license when you can sell 3!
I mean, I think backups and snapshots that actually work is something that most people would benefit from?
There are few pieces of software that are more universally disliked than Windows Update, and while I cannot speak for everyone the main reason I hate it so much is because it's often automatic and I've had it make my computer unbootable multiple times (and it's actually a big reason I ditched Windows completely like sixteen years ago).
If there were full filesystem snapshots and a utility to fully restore from those snapshots, I think a lot of people would be more willing to do the updates. It wouldn't be hard for users that have any tech skills at all; do what NixOS does [1] and on boot allow the user to choose an earlier snapshot to boot from if anything gets toasted.
Windows does have System Restore, but as far as I can tell that never has fixed anyone's problems at any point in time, along with the "Automatic Repair Tools" which has become a running joke in my family.
It just bothers me, because the FOSS community has solved this, a long time ago no less. Back in 2012, I used btrfs + Snapper whenever I would install a "risky" driver like a GPU or Wi-Fi chip, or whenever I did an update. When I did break my video display, restoring from the snapshot took like ten minutes, I'm able to get back into my system and fix the issue, without having to do any major surgery to the internals of Linux. I've never tried it with FreeBSD and ZFS but I suspect it's a similar process.
Microsoft is a trillion dollar corporation, they have a higher market cap than the entire GDP of NYC and Singapore combined. Why exactly can't they just augment ReFS and copy Snapper or Time Machine (with APFS)? They have infinite funds, they have access to Dave Cutler and until about a year ago Leslie Lamport, they already have a CoW filesystem and they could easily afford a license to any version of ZFS if ReFS wasn't up to the task.
[1] I know NixOS isn't doing filesystem interface, but the overall principle is the same in this instance.
And on a client machine, it's of much less importance overall (to you, it may be super important and I don't want to discount that). And on the server side, that's why we have n+2 failovers. No single machine of importance should ever be a point of failure... I realize that's not always reality but it's more or less Microsoft's position; after all, why sell one Windows Server license when you can sell 3!