Well you don't have to use interns to get production code. If that's a concern, they can be put to work doing something outside the main code base, like perhaps writing tools that speed up some tedious process. There's probably opportunity for OpenBSD to get something out of GSOC, but if the standard is "shippable code or bust", well, no wonder they're disappointed.
(i.e. less emphasis on experimenting and more emphasis on quality and being accepted by upstream, perhaps by assigning "smaller" tasks)
>without an eye to maintainability that later rots in tree and makes everyone else's job worse.
...that's the whole point of the mentors. IF that's what OpenBSD is getting out of GSoC, it's the mentor(s) fault, not GSoCs.
They have no issues attracting devs via other channels.
GSOC promotes dumps of low to mid quality code that then goes unmaintained. After all its focused on relatively junior programmers doing it for a nice stipend as part of their school experience for a summer (or so).
GSOC is just a bad fit for the sort of quality and maintainership OpenBSD demands from its codebase and contributors. They have had a lot of years of experience with academic code mostly written as a proof of concept without an eye to maintainability that later rots in tree and makes everyone else's job worse.