cmos relaxation RC oscillator was found to be frequency stable when i swept temperature ranging from -40 degree to 100 degree.when i plotted propagation delay of the cmos inverters which i used for this design,it was found to be varying with temperature sweep.Why the change in propagation delay not affecting the frequency of operation ? In the case of ring oscillator ,propagation delay is varying much and it is affecting the frequency ..what point am I missing here ? please help

There could be some kind of cancellation hapenning. I suppose you are using comparators, could be possible that comparators(I/C or RC kind of osc's) are slowing down at 100c and your reference current is increasing at 100c,effectively ending up compensating each other.

Faster is a relative term. As the Osc frequency reaches say tens of Mhz Comparator delay begins to play a role (say 20Mhz or more).Basically if without applying any cancellation techniques, you end up getting very good freq accuracy, then you are probably seeing some cancellation (eg:comp delay more at 100c causing output freq to decrease and Iref  increasing at 100c). One method to isolate causes in such a scenario would be to plot sc freq vs Temp for atleast 5 temperature points and you would see the freq variation peaking at a point other than 100c.