Because it may take up to 2 minutes for the spider run to stop completely and the Unit's resources to be released.
This issue is often observed in projects running very fast jobs -- with runtimes around 1 minute. The job may appear as finished, but its resources are actually still being released, i.e. the Unit is still not available.
We are aware of the issue and we have it in our internal roadmap to improve this process.
In the interim, we recommend aiming for longer running jobs (within reasonable limits). For instance, it may be possible to aggregate a number of extraction tasks into a single spider, rather than scheduling a multitude of spiders with one task per each.