Finally some useful actionable feedback! Thank you @besentv. I'll bring back the `DECLSPEC_HOTPATCH` where I removed it, replace the `%llx` + casts with `%Ix` (casting to `ULONG_PTR`) for pointer-sized masks, copy the cpu mask while holding the NUMA lock, and tidy the contiguous-remap finalization. I'll also clean up the thread-affinity debug traces by removing the unnecessary if/else statements to factor the common finalization/trace and avoid redundant conditions. I will take this opportunity to move those changes in ntdll since everyone is so adamant on having them there (for good reasons).