> Surely losing less than 100MB of data from a drive should not render hundreds of GB of perfectly intact data unreadable!

Considering btrfs doesn't have a working RAID implementation, screwing up a drive would lead to the volume being unusable, yes.

Raid 0 and raid 1 absolutely work well on btrfs, I've been running in production at scale for more than a decade.

(Raid 5 and 6 are a different story, but frankly I think it's silly that modern systems talk about them at all, instead of general purpose erasure-coding.)

Symbology devices ship with btrfs based raid by default.

Synology's RAID is based on a combination of Btrfs and MD, it doesn't employ Btrfs' multi-device support at all.[0]

[0] https://kb.synology.com/en-nz/DSM/tutorial/What_was_the_RAID...