v22 is live btw.
Part of the difficulty with predicting is that CAD software, generally, is highly dependent on many vendors. For example, we were going to release a few weeks ago, but we got an unexpected update that fixed a lot of great bugs and long-standing issues with HOOPS. 100% worth including and delaying a bit for. Cool, ship it! But we can't - HOOPS touches so many things that it is perfectly likely that some other thing broke - back the drawing board for testing the full product. During that, we got a nice update from DCM - cool, ship it - but it's always trust but verify. We had to test a bunch of sketch related functionality again.
If we made everything ourselves, we would have the advantage of greater predictability with the disadvantage of shipping once every six years. Different software teams with different products and different dependency levels can predict the future with quite varying degrees of certainty.
It's also not hyper critical for us to make an exact ship date - the overwhelming feedback we've gotten is "ship it when it's ready" - no one gets angry if it isn't shipped on an exact day predicted 3 months in advance. Everyone gets angry if it's broken.