I hit this with procps. (a package with ps, top, vmstat, free, kill...) It was horrifically demotivating, helping to end my involvement as the maintainer from roughly 1997 to 2007. (the other big issue was real life intruding, with me joining a start-up and having 5 kids)
I had plans for command line option letters, carefully paying attention to compatibility with other UNIX-like systems, and then Red Hat would come along and patch their package to block my plans. They were changing the interface without even asking for a letter allocation. I was then sort of stuck. I could ignore them, but then my users would have all sorts of compatibility problems and Red Hat would likely keep patching in their own allocation. I could accept the allocation, letting Red Hat have control over my interface.
Red Hat sometimes added major bugs. I'd get lots of complaints in my email. These would be a mystery until I figured out that the user had a buggy Red Hat change.
Patches would often be hoarded by Linux distributions. I used to regularly download packages and open them up to look for new crazy patches. Sometimes I took the patches, sometimes I ignored the patches, and sometimes a wrote my own versions. What I could never do was reject patches. The upstream software maintainer has no ability to do that.
The backlog of unresolved troubles of this sort kept growing, making me really miserable. Eventually I just gave up on trying to put out a new release. That was painful, since I'd written ps itself and being the maintainer had become part of my identity. Letting go was not easy.
Maybe it had to happen at some point, since I now have more than twice as many kids, but I will be forever bitter about how Red Hat didn't give a damn about the maintainer relationship.