Safely cancelling jobs

Hi,

Is there a way to safely cancelling jobs without needing to know if they have started, running or finished? I do multiple simulations in parallel, and many of them are finished around the same time. Some of the results become redundant after processing the others, but calling e.g. `async_res.abort()` on the redundant ones quite often freezes the execution (probably the assertion of the job not being finished passes but in the meanwhile the result arrives). So currently I can't abort them because of the chance of freezing the execution.

Start by tracing the async_res.abort() lifecycle and how job completion is handled when several simulations finish together. Reproduce the race between aborting a redundant result and its completion, then define safe behavior for jobs that are pending, running, or finished without freezing execution.

Do mô hình lập chỉ mục viết ra từ nội dung của issue.