@joebonrichie Have you experimented with progess tick values of 4-10ms yet?

How do we know that two seemingly unrelated triggers won't each write to the same file, meaning the trigger outcome is not deterministic (because the output might be different than if we knew that one would always come after the other)?

How do we know that two seemingly unrelated triggers won't each write to the same file, meaning the trigger outcome is not deterministic (because the output might be different than if we knew that one would always come after the other)?

In a nutshell, we don't. We have no mechanism for recording what files each trigger writes to

How do we know that two seemingly unrelated triggers won't each write to the same file, meaning the trigger outcome is not deterministic (because the output might be different than if we knew that one would always come after the other)?

In a nutshell, we don't. We have no mechanism for recording what files each trigger writes to.

In the context of tx triggers, we arguably should, as this is stuff that is under our control and will become a part of a tx via the postblit stuff. BTW, this could in fact be one of the reasons that Ikey chose not to parallelize tx triggers initially.

In fact, doing the tx triggers serially would allow us to create a "before" list of files and an "after" list of files for each trigger.

In the context of system triggers, this seems a bit pointless, as the tools we call should handle not writing to stuff they have no business writing to.

Definitely a valid concern, but surely outside the scope of the this PR unless there are tx triggers writing to the same file without holding a lock on the file.

Definitely a valid concern, but surely outside the scope of the this PR unless there are tx triggers writing to the same file without holding a lock on the file.

Well, the solution we have now is arguably the more correct one, while this PR potentially sacrifices that correctness on the altar of speed. Hence the question.

Converting this to draft until we have a better handle on the correctness story.