Code review comment for lp:~jameinel/bzr/2.1-peak-mem-tweak

In terms of code quality, this looks good. You've commented out an assert of internal_node_cache size post calling clear_cache() though. Maybe you want to grab the size beforehand and check it's unchanged afterwards? Or maybe that line can go altogether.

I'll run some usertest benchmarks in coming days once you're ready for that. As we measure more, we may tweak things further I guess. Even so, I can't see any faults in this patch and it's a step forward resource usage wise so I'm happy to see this land.