Suppose you have a video converting app. Further suppose your headline new feature is converting 2x as fast as the old version, pretty much the holy grail of video conversion software.
In the MAS you have a set of bad choices:
A) Release the upgrade free to everybody. Great for them, bad for you.
B) Make the new encoder an IAP. Problem is, now you are maintaining "old" code that you may not want to maintain anymore. You may need to wrestle with the old encoder to get it compile for new versions of the compiler, etc. You have to fix bugs that occur with the old encoder and new, unanticipated changes in OSX. If you're still compiling and uploading it every month you can't plausibly claim that it's no longer supported.
C) ship the new encoder, but cripple it so it's slow until users upgrade. This seems capricious. Also, you can't get exactly the performance of the old encoder; you'll either overshoot or undershoot. And the nature of the MAS is that user's can't effectively downgrade, so they'll be SOL if you're wrong.
D) Remove the app from sale and submit a new app. You're wiping your slate with links, app reviews, etc. You also lose customers, who don't find out about your shiny new version. OmniGroup pursues this strategy, and they lose a lot of revenue from me (an avid user) because I completely miss certain upgrades because I don't hear about them. I can't imagine how it is for marginal users.
E) Don't update the app after all, because in spite of the fact that 2x video encoders obviously add value to the world, the MAS adds too much friction to make it a viable business. I would argue that the majority of iOS and Mac apps face this sad reality at some point in their lifecycle.
I submit to you, that if you manage to write a video encoder that is 2x faster, that should be the hard problem. The hard problem shouldn't be to figure out how to structure your upgrade pricing. But with the MAS, it is.