Isn't that a bit of rewriting the history? Originally the roadmap was that these kinds of features would be added based on the mythical block pointer rewrite. Eventually BPR and thus all features requiring it were put on ice.
Had Sun remained the powerhouse it was, the BPR would probably probably have eventually gotten done, but with the smaller resources devoted to ZFS development now vs in the Sun days, it's gotten kicked way down the priority list.
That is an insanely time consuming operation that also increases the risk of data loss as you have less redundancy during the expansion.
An enterprise ought to just migrate the entire pool/content to somewhere else and redo it completely.
Expanding by replacing with higher capacity drives just feels like a hack and afterthought that only hobbyists would actually make use of.
If the old device is not faulty then the replace operation will keep both old and new in the array until it is complete.
> An enterprise ought to just migrate the entire pool/content to somewhere else and redo it completely.
That's also an option with send/receive, but that involves downtime. Drive replacements can be done on a live system. Server grade hardware generally supports hotplug.
Which you probably should.
ZFS was not philosophically opposed to those ideas. They were just dropped at some point.