Some further observations.

I have now changed the intra-sync interval to 1 hour (the 10 minutes was for
initial tests only).

It seems that the process dies 1 hour (that is, the length of intra-sync
interval) after it has been "disturbed" by the command-line tool, irrespective
of when the next sync was due. And that sync does not occur. 

For example, if a sync took place at 13:15, and I use the command-line tool at
13:40, the 14:15 sync will not take place, and at approximately 14:40 the
process disappears from the process table. Can it be, that the 1-hour timer is
re-started from zero after interaction with the command-line tool, and when the
process wakes up from it, the sync time has passed, which confuses the
algorithm, leading to the crash?