If this isn't due to latent memory pressure, it's usually because _something_ in your code is trying to allocate a gigantic amount of memory due to a bug. Think a 1 billion x 1 billion numpy array, or a billion element list.