There is a surprising amount of excellent computer science research that is done as a private hobby where no attempt is made to publish it. An important aspect of this is that the goal of the research is often to solve the problem. Publishing it is irrelevant to that objective and a tedious chore that can be readily dismissed if it is not your job. It may not be a satisfying answer but in my experience "I have better things to do with my time" is a primary reason to not publish amateur research.
This naturally creates two problems. The first is discovery amidst the noise of non-serious research activity. To find it you have to be aware the research is happening and you literally have to talk to the people doing it. One thing NSA used to (maybe still does) do very well was finding non-academic researchers doing interesting theoretical computer science work that was never going to end up in a journal. The second is that there are a couple domains where there are decades of accumulated research where almost none of it was published, so it is difficult to bootstrap yourself into the state-of-the-art because so little of it was written down. There is also the reality that some traditional sponsors of non-academic computer science research have asymmetric advantage as an explicit objective, which benefits from non-publication.
This naturally creates two problems. The first is discovery amidst the noise of non-serious research activity. To find it you have to be aware the research is happening and you literally have to talk to the people doing it. One thing NSA used to (maybe still does) do very well was finding non-academic researchers doing interesting theoretical computer science work that was never going to end up in a journal. The second is that there are a couple domains where there are decades of accumulated research where almost none of it was published, so it is difficult to bootstrap yourself into the state-of-the-art because so little of it was written down. There is also the reality that some traditional sponsors of non-academic computer science research have asymmetric advantage as an explicit objective, which benefits from non-publication.