Off-peak, the autoscaler removes nodes and our Spark driver got evicted mid-run. How does scale-down choose its victims?

Scale-downers mark a node removable only when every pod on it can leave safely. Your batch workload died because nothing marked it special. The mechanism is a filter cascade, and the fix is priorities, PDBs and annotations.

Grounded in researched DevOps, SRE and platform engineering interview loops, written to a senior-engineer editorial bar, and never padded to hit a word count.