6.2.0 has no manual to download

checksums and, more specificially, healing, rely upon BtRFS's (or md5 if that's the layer it's happening in ROS6) knowledge of extra copies of blocks and/or parity information at the block level with which it can attempt to recovery from checksum errors. That protection, in BtRFS, occurs even if I choose to disable CoW on a folder (like the one I keep my VM's in). Can I still snapshot that path? You betcha! I do it every single day in my Linux environments. And, yes, the snapshot will result in CoW as data diverges from the snapshot... but once the new blocks are written for the +COW path(s), it stops doing additional CoW on those blocks.

Yes, that can get messy after a long while... but it's part of the mix to be aware of when tweaking around with filesystem behavior. :)

Perhaps some details will diverge from this in terms of how Netgear implemented it (bringing md5 and BtRFS together), but without documentation that's hard to say. I can confirm how it works in a pure BtRFS environment.