That would potentially be one of the problems. Having to run undo countless times while experimenting would be another.
ok, but in the end it is basically just data ping pong under the hood. for undo you could make some timeout i believe, meaning when there is no further change that the last value gets tossed into the undo/redo memory and not the numbers in between. and when you change a value field that would also finish the last value field. i mean i see now that there is something to consider and these are just dilettante ideas of mine, but not impossible maybe and if it could lead to a better/faster/easier experience then why not?
@stevebaer as somebody has just asked about this bug here i wonder if the bugtracker you mentioned is different to what i discovered and if so if it is actually public or if not if it (accidentally) remained closed?