Distortion is probably one of the DSP tasks where oversampling is most beneficial because it introduces a lot of higher-frequency harmonics. At the standard sample rate of 44.1 kHz, the Nyquist frequency is 22.05 kHz. Any harmonics above this point will fold back into the audible range as ugly aliasing noises. When you go to higher sampling rates like 96 kHz+, the amount of energy above the Nyquist frequency is significantly lower and the stuff that is there will mostly fold back into inaudible frequencies. You can easily hear the difference by using Scream 4 at different sample rate settings. You can also see the amount of aliasing on a spectrum analyzer.
I think the aliasing issues should be less significant on synth oscillators unless you are using unfiltered oscillators with a lot of high frequency harmonics (like sawtooth) and playing on the highest keys. Of course, if the synth uses any sort of clipping/filter drive, etc. then it will probably introduce some aliasing from that. Aliasing in synths is a bigger issue in FM because of the large amount of high frequency harmonics. Try playing some FM patches at two different sample rates and they might sound completely different.
I would recommend that Propellerhead implement oversampling switches on every device, but particularly on devices that use distortion algorithms like Scream 4, Pulveriser, Echo, Line6 amp models, etc. The CPU hit should not be a problem because most users can disable the switches during tracking and only switch it on for select devices when they render the audio.