Make sure that tasks do not replicate themselves

[sveneberth]

I think the problem has existed since ViUR 2 ...

I experience it more and more often that with long tasks (many chunks, over hours / days), the task is duplicated and executed several times in parallel. On the one hand, this naturally leads to more resources being consumed, on the other hand it leads to transaction collisions.

This can actually only happen if an error occurs between the call of the next chunk and the completion of the task (network error, timeout), so that the next chunk is queued but the last chunk is evaluated as failed.

Some mechanism is needed here that recognizes that a task is already one execution further.