Undiagnosed memory leak

Apologies for the vague title, but we're seeing an issue with a memory leak on a customer's instance, and we're having trouble finding the root cause.

The issue exists on 10.6.19 and 10.6.22, and occurs with both the system malloc and jemalloc allocators. The memory problem builds very slowly, but requires a restart roughly once per week to free up the memory.

I've attached the output of a few memleak reports. If there are any additional flags we need to pass in order to capture more information, let me know, and I can rerun the tool.