This is quite useful if one task failed (for example when a worker times
out) and you want to re-queue it without cancelling any currently active
tasks.