Current git loading tasks can fail to be ingested with a [1] 'pack file too big' message.
They are then disabled by the scheduler after a number of tryouts (3).
Those could be redirected to another queue whose consumption can happen only with large enough workers [2].
(with proper setup like pack file limit set to 32g or something).
[1]
```
raise IOError(
f"Pack file too big for repository {origin_url}, "
f"limit is {size_limit} bytes, current size is {cur_size}, "
f"would write {would_write}"
)
```
[2] worker17-18 currently