I'm not sure how I feel about this, as it changes the definition of "interval". It could be surprising to some people. This change increases throughput, but also adds DB load.

It's also arguably a breaking change.

Fair points; on the other hand, I don't think it's common/expected for a task queue to throttle its throughput to keep resource use under control. If a user really wants this, they could add a time.sleep() themselves, either in the task, or in a post-execute hook.

Change looks good now! Although I'm a little concerned about it being a breaking change. The help text for --interval mentions checking for new tasks. I can't think of a better name than "interval", but the text should be updated to match its new use.

Although I'm a little concerned about it being a breaking change.

That's a good argument for the next release being 0.4.0 :)

but the text should be updated to match its new use.

How about The interval (in seconds) to wait, when there are no tasks in the queue, before checking for tasks again?

The interval (in seconds) to wait, when there are no tasks in the queue, before checking for tasks again?

Works for me!

the next release being 0.4.0

It probably will be. I've got some other bigger changes in the work which warrant a larger release bump.

Alright; I've updated the sleep condition and the help text for the CLI.