After spending a decent amount of time looking at a customer issue on 6.0 with heavy TCMalloc spinlock contention, I had them run on 8.0, but the spinlock metric was missing, so I could not determine whether the new allocator was any better.

That metric was incredibly valuable for quickly identifying allocator contention, and I think we need to find a suitable replacement for 8.0.

In particular, we need a metric that tells us how many operations are bottlenecked on the allocator.