daniel_sahal
Databricks MVP

It could be due to the caching that may use some amount of memory when you're reusing cluster.

Simply try increasing your memory and/or optimize your code a little bit.