I wonder if there are any people actively working on this
I'm not aware of ongoing work on this topic. However, it's more or less feature-complete from Platform POV (there may be bugs to report and fix, but nothing new is planned nor seem necessary at the moment).
If you're more specifically things about JDT, I can't tell. My previous impression (that may be outdated), is that concurrency wasn't visibly improving JDT Build time and was introducing some issues. So the next steps would be 1. verifying that JavaBuilder can benefit from concurrency (eg by relaxing a few things in its code and making a few test to compare performance, ignoring potential glitches to be fixed later) and 2. if it's profitable, identify the issues concurrency can/does cause on Java Builder and fix them.
and if there is any design documentation?
I think the EclipseCon presentation is the most detailed design documentation, after the source code of course.
what the intentions are
Faster workspace builds when possible ;)
I understand this is all experimental
It's not that experimental, it's officially in here and can be used with same level of support as anything else. We're aware of at least a few people who happily use parallel builds as their builders do invoke things that can be easily run concurrently. It's however not enabled by default in the workspace because there has been no urge from it.
One thing is that with language servers and so on, the build step in the IDE has become less important: most projects are just not built by/in the IDE, they're processed by the language server and the debug adapter to do what's necessary. That's the reason why my personal interest in optimimizing workspace build did reduce. And as for server-side (eg in JDT-LS), this has not been considered yet as focus is not so much on this level of perf optimization at the moment.
but it would be nice to have some discussion group somewhere to see if this can be moved forward.
Can you please elaborate about what you mean by "this" and by "forward"? ;)
