Games are a good example. One might want to publish a game and then work on the next game, not go back to the first game again and add dynamic permission prompts for the accelerometer or recompile with the new SDK or whatever. But someone also might want to play Space Grocer I before Space Grocer II-X to get the whole story.
The fewer breaking changes there are, the lower the burden of an app being "supported" is. Someone might be willing to recompile the app every couple years, or add a new required argument to a function call, but not really able to commit to re-architecting the program to deal with completely new and now-mandatory concepts.
Even on software I actively work on that is "supported" by me, I struggle with the frequency of e.g. angry messages demanding I upgrade to new and incompatible versions of Node modules. Time spent porting to new and incompatible versions of a framework is time not spent keeping the app worth using.