OK - this one bothers me more than it should. I understand library problems with Galileo timing, and missing AVR headers (different architecture), but you don't go changing the Arduino language for no good reason.
There is no reason for the Galileo implementation of random (and randomSeed) to be different from Arduino 1.05.
After a bit more research - this looks more like a sin of omission - not commission.
Intel did not provide a random or randomSeed, so the posix definition from stdlib took over.
Back to being not bothered - just add it to the bug list... We really need a conforming implementation of all Arduino functions.