Nah, thats due to it being BTRFS not being COW.
Netapp filers for example are copy-on-write (WAFL). ZFS is COW.
COW doesn’t mean it copies a block to write, it just means the new write goes to another part of the disk. COW is by no means inherently slow - and if you’re doing things like parity RAID, COW will actually speed things up as it avoids the read-modify-write cycle to update in place (and also avoids the write hole problem).
Oracle basically bought SUN for ZFS, so the whole “disable COW on databases” for BTRFS is just an example of BTRFS being broken, not COW.
Some people will argue with you, but they’re wrong - out of the currently available options this is indeed the case. ZFS has some trivial (in today’s landscape) overheads on top of regular RAID, but it actually returns the same data you put into it - reliably. Which is kinda the point …