I found that I can make my machine spike in memory usage with just this:

import matplotlib.pyplot as plt
from matplotlib.ticker import MultipleLocator

fig, ax = plt.subplots()
ax.set_yscale('log')
ax.yaxis.set_major_locator(MultipleLocator(100))
plt.savefig('test.png')

When I say huge, I mean I see memory usage go up and I kill it before I end up with an unresponsive system.

Removing either of the log scaling or the MultipleLocator causes the huge memory spike to disappear.