@amcasey,
Thanks for your contribution as a Microsoft full-time employee or intern. You do not need to sign a CLA.
Thanks,
Microsoft Pull Request Bot

Fixes #18255

Some thoughts about massive projects

• I'm worried about the magic limit 10 not being enough to protect against the buffer being blown. I think the project we were investigating might have had less than 10 projects. Perhaps we could have a limit on total size of in-progress requests: we could use a heuristic for the size such as request.fileNames.length?

  • Edit: I'm having second thoughts about this constant limit approach altogether. We (I, at least) don't really have much insight on the underlying buffer yet, and I'm not confident it's going to be the same size across different hardware, Node releases, etc. At the very least, I think we should handle the case where we still fill the buffer, despite our mitigations (instead of just getting stuck). I'll try to understand how we could do this a little better tomorrow.

• On a related note, do we still queue a typings request for a project we've disabled as a result of the 20MB limit? If so, it should be a no-brainer to quit doing that.

Failures are lint. I'll update.

@minestarks Are disabled projects flagged in some way?

languageServiceEnabled

It does not appear to be skipped at present. I'd like to fix that in a separate PR though.

@amcasey can you port this to release-2.5

@mhegazy we're just waiting on final validation from an affected user.