GEN2 dataflows are super expensive. We are seeing them kicked off at random times, without any explanation. If they fail after an hour we want the fail to only cost us an hour. If we want to schedule an additional redundant refresh of a dataflow, then we will do so ourselves.
Microsoft appears to be benefitting from kicking off long-running dataflows on their own initiative. I think the result is that they can charge us lots of CU's for it and it will often be unnoticed.
See below that I have a dataflow scheduled at 7:30.
However in the history for this dataflow, we see two distinct instances being executed this morning.
The second one is not in the schedule.
Is this a "known issue" in dataflow GEN2? Will the PG (and Mindtree) see this as a bug or a feature?