Is there any tuning we can do to get the size of the compression context down, since we know the maximum size if the input will be 16MB?

Just some optional performance suggestions