The section at the beginning that hypothesizes the existence of security tags on registers (or rather, on architectural <-> physical register mappings), based o

The section at the beginning that hypothesizes the existence of security tags on registers (or rather, on architectural <-> physical register mappings), based on observed performance when reading an "unowned" register, is super interesting.