This worked in that it removed a fixed cost, although it's not a huge proportion (~3-10%).
Going through the process brought out some more low-hanging fruit for optimizations.
(And once other costs are reduced, the fixed cost here will represent a larger percentage 😂)
I've got an idea that I think should majorly speed up breakpoint recovery that I'll be trying out this week.
All going well, this should mean dense inspections cause less slowdown. The effect should be particularly noticeable for multithreaded user code.