The non destructive group transforms are great, in fact they are one of my favourite features. But the destructive versions do things that they don’t. Destructive can be better for certain applications of randomness, for example, because it isn’t ephemeral and reevaluated on the next transform. Also the destructive transforms can be performed directly on the module without external signals.
For example, I have been working on a technique for my next video where I have a table I’ve set up with different scales in different parts of the table. The group transforms would be great for jumping one partition the table in response to external gates and so on, but there are a number of operations that are useful to just grab by hand in performance. Jumping up and down parts of the table (ie up up, invert, down down), gradually jittering so that the resultant random walk leaks into adajcent partitions, or randomly re-seeding a pattern with a combination of rand and addition. These are all useful things to do destructively by hand, and being able to set each transform up before hand and then recall them would add a great deal of fluidity to improvised performance.
I certainly wouldn’t want this to replace the group transforms, because these are better in other ways.
I guess the way I’m thinking about it is I want destructive transforms as quick ways of setting up and modifying the sequence, whereas the group transforms are for setting up variations on top of the sequence once it is set up.