Hi Tony,
Personally, I think only projects proposed by potential mentors should be
considered at all, and that the documentation should state that clearly. I'm not aware of any GSoC projects where the student came up with the idea on their own and then executed on it successfully - with the exception of projects where the student is an established MediaWiki developer who happens to currently be in college, but that's obviously a special case. It's just not reasonable to expect that someone from outside the WMF/MediaWiki community would be able to come up with a project that (a) makes sense, (b) fits within the current development roadmap, and (c) is of the right scope for a GSoC/Outreachy project.
You are right, and a student proposing a project few days before the GSoC project cannot make (a) sense and (b) fit anywhere. This is exactly why we are trying to solve. With such a program, the student gets in cool with the community a bit (as having a few months headstart), and we might even end up having students who might be able to understand the community to propose something that makes sense ? I know there is whole lot of optimism in there, but there can be chance. Like in my case, my GSoC 14 project on VERP was not a featured project, and strangely one day Nemo Bis talked to me about this while roamin around in #wikimedia-dev. I just think better connections and understanding of the community might even bring in better projects (too much opitmism there).
I still think coming up with project ideas should basically be left to potential mentors. In your case, it's not that you actually came up with the project on your own, as in you saw a need for better email bounce handling or whatever it was; it's that this was a known problem, but no one thought to make it a GSoC project until you talked to a developer. Which is a problem, and I think it stems from the same apathy that has led to not enough WMF mentors. But I hope that there's a better solution for it other than essentially requiring potential students to become detectives, trying to find interesting coding challenges that no one has proposed for GSoC etc. Maybe the solution is for you and others to do this work yourself - talking to MW/WMF developers to find more tasks and drum up enthusiasm among potential mentors - essentially what you did before, but now as an administrator and not a potential student.
-Yaron