This is still work-in-progress, and slated for the 0.15 release. It's an absolutely necessary performance fix in case get_results does any significant work (like CoM with regression=1), and is especially useful in combination with rate limiting visualization results, also pushing us one step further towards not having to manually touch the partition sizes together with LiberTEM-live (we can just make them smaller by default!).

Main TODOs:

• How to make this thread-safe - we need to prevent the delayed get_results and the merge function from running concurrently on the same data / buffers, basically.
• Needs documentation on how to safely access the results of run_udf_iter
• This slightly changes the observed behavior, as seen in the failing test cases. I think this should be OK and the tests and the docs need to be updated, but some input would be welcome! This is mostly an issue if get_results has observable side-effects

Fixes #1596

Contributor Checklist:

• I have added or updated my entry in the creators.json file
• I have added a changelog entry for my contribution
• I have added/updated documentation for all user-facing changes
• I have added/updated test cases
• I have included the rebuilt production build of the client (only if changes were made to the GUI)

Reviewer Checklist:

• /azp run libertem.libertem-data passed
• No import of GPL code from MIT code