Right now, if the $sample query (used to create the DRDL schema) fails, it'll re-run the query once again, forever. The only place when the error is accessible is in the "Profiler" tab of a given node. In our case - and I believe it's common - that it was a secondary node, which "Profiler" is far less accessible than the one for the primary node.

We learned it the hard way, as our cluster was silently maxing out one CPU core for over a week. After a lot of debugging, it turned out that one of our views has had a problem ($arrayToObject received a null key).