I happen to encounter something similar.
it's probably because you are just `explain` it. when you actually `run`
it. you will get the final spark plan in which case the exchange will be
reused.
right, this is different compared with 3.1 probably because the upgraded
aqe.
not sure whether this is
Spark Devs,
what's the potential security issue to reuse one spark session for
running spark-hive queries, spark jobs ? suppose the session usage is
isolated to different time chunks , and some clean up is done before each
usage (for example hive cache is recreated before each usage).
thanks.
