Why in the hell FreeBSD changed semantics of RFMEM flag in rfork(2), compared to original call in Plan 9? This misunderstanding cost me countless hours wasted on debugging...