Memory leak from looping through netCDF files only in notebook cells

[djgagne]

I am running into a memory leak when I loop through a set of netCDF files, read their contents and plot them, but only when I run this loop in a Jupyter notebook. The same memory leak happens when I use xarray, netCDF4-python, or h5py as the underlying library to read the package. I have tried adding explicit del and gc.collect commands in the loop but still see the same amount of memory growth. I also put the loop within a function, so all variables should be garbage collecting and do not have global scope. Does anyone have any idea what might be going on under the hood in Jupyter that could be causing memory retention/memory leak that is invisible to the Python garbage collector?