The scheduler does not take full snapshots on the production server and stops mid-through.
After some investigation, it seems to be caused by an OOM error due to the size of the data returned from SPARQL queries.
The scheduler has been disabled until the issue is fixed.