I am having issues with my research notebook stability when pulling in data for model training.  While the code for pulling the data is well tested and usually works fine, especially early in the morning, it often crashes after 60-70% of the data has been loaded.  This is extremely frustrating as it often takes 40-50 minutes to load the data.  While the dataset is not large (25MB when stored as parquet), I already have to chunk the data to keep the notebook from crashing.   Also, I have noticed that it seems to crash more often in the afternoons. It also seems to run much more slowly in the afternoons, but I haven't timed performance to say for certain.  I am assuming because the servers have a higher load?  Is there anyone else having these issues or found a solution?