It’s true that you have plenty of power with a non-UW MD. It’s also true that a UW model is absolutely worth having.
The internal synths are an endless source of sounds. Based on my past 9 years, I can say with confidence that you could explore them every day for the rest of your life and never get bored. Early adopters launched a version of our “science labs” where they’d make a kit out of 16 of the same machine and tweak each one to a different flavor of unrecognizability.
Combine this with its other strengths—16 freely assignable LFOs, control machines, a powerful sequencer—and you’ve got lots of capability.
That said, the UW engine is something I’d never want to be without.
It’s a shortcut to immediately useful sounds if nothing else. Feeling lazy or impatient? Want to just bang out a beat to go with that synth melody you just wrote? Load up some new samples and you’ve got instant gratification. It also opens up the MD to single-cycle waves, which are great for basslines, and melodic samples, which can turn the MD into a more or less complete workstation once you learn it.
Personally I’d want a UW if it only offered the resampling feature with no sample loading at all. I’ll spare you, but I could talk your ear off.
Some people gravitate toward one approach more than another, so of course there’s no single right answer here. But put samples and the synth voices together, all interacting in a single environment, and you can go a mind-boggling variety of directions.