Loading Into Columnstore: Avoid Trickle Loads

I'm going to tell this one in story format, so here's the short of it up front: tl;dr --- Clustered columnstore indexes don't like the combination of wipe-and-replace with multi-threaded trickle loaders.  Avoid that pattern. The Setup In the olden days, we had a large fact table with a standard clustered index and some standard non-clustered indexes.…