Would the use of compression and/or encryption on the zfspool/lxd dataset cause any problems for LXD? Or other downsides? Or is it more ideal to create via: “sudo zfs create zfspool/lxd” ?
As far as I can tell, the general consensus with zfs seems to be “use the defaults”.
Nobody’s going to argue against ashift=12 though (it should be default these days), and compression is almost always a good thing: the (small) CPU hit for compression/decompression is more than offset by the reduction in I/O.
Personally I don’t see any benefit in encryption when the encryption key is stored on local disk on the same system, but that’s for your own situation and risk profile to decide.
All of this is transparent to lxd/incus: it’s just another workload, and containers are just processes.