Tested both now and the first one varies with +- 0.12 rpm which is fantastic and the second one are absolutely still as its a long, (no decimals).
I know what you mean with the integrating math and I will have some small integration, so it wont affect the RPM acceleration.
So a overflow test isn't necessary when checking passed time?
what if the micros reseted between assigning rpm_timeold and the difference test, wouldn't that be negative or 0(unsigned)?
timetemp = micros() - rpm_timeold;
if( timetemp >= 0 ) { //if micros not resetted (overflow every 70 minutes)