These are nifty data structures that can quickly search the space of equivalent programs.
update: oh, I just remembered why it _is_ an issue for modern processors, Massalin leaned very heavily on the general purpose double compare-and-swap that was present on some generations of the 68k architecture, and x86-64 only has double contiguous compare and swap
It constantly trys to do the most things with the least things. Take all the shortcuts to achieve the same outcome, often misconstrued with laziness, quality deficiency, or hard to understand.
For example, something as simple as timing may superoptimize and minimize instruction sets. Using one resource to do all the things needed at once versus calling the resource multiple times at different times shrinks the overall program a lot.