I’d shy away from deduplication except in a very limited set of circumstances and those I would research on the old forum. I seem to recall someone using some very fast SSD storage (optane?) in a pool for a multi-VM hosting application where ZFS deduplication was successful at paring down drive needs dramatically.
But the key being the use of Optane, which is insanely fast, and a very specific use case. I would treat deduplication just like a sVDEV, ie an awesome tool for the right use case but also a tool to wield with great care due to the potential downsides associated with using it incorrectly.