Small companies go under with startling frequency, and even with an ironclad contract, there's often nothing you can do but take them to court when they've gone bankrupt. Unless you've specifically contracted for source access, you're completely SOL. Profitable niche companies lose interest too, and while you may not have the same problems if they sell out, you'll eventually have very similar problems that you can't do anything about.

Consider any of my dozens of little OSS libraries that a handful of people have used, on the other hand. Maybe I lost interest a while ago, but it's pretty well written still (can't have people judging my work) and when you realize it needs to do something, or be updated (since things like dependabot can automatically tell you long after I'm gone), you're free and licensed to go make all the changes you need to.

I think you see highly specialized software being run by enterprisey companies because that's just business, not because it's better. It's easiest to start in a niche and grow from there, but that holds true with open software and protocols too. Just look at the internet: used to share research projects between a handful of universities, and now has grown to petabytes of cat gifs. Or linux. Started out as a hobby operating system for a handful of unix geeks, and now runs 96.3 percent of the top 1 million web servers.

It always starts small and gets better if it's good enough. This goes for OSS and companies.