Spark application has different runtime on the same amount of executors and memory

Question

I have a Spark application scheduled to run at different times of day. I ran it in a busy production queue for a few days and then switched to an empty dev queue. I monitore