I, like many of you, presumably, have a number of plugins that can oversample up to 8x. Is there a point of diminishing returns? 2x? 4x? 
 
I'm away from my studio right now, so I can't A/B, but my guess is that I'd have a hard time telling the difference 4x and 8x, if not 2x and 8x, on an individual plug. Nonetheless, it seems like oversampling is probably helpful cumulatively, given that I sometimes have as many as 5 or 6 plugs on a track. I'm just wondering if there's a point after which it gets silly, sort of like recording at 192 kHz, though I'm sure there are folks who swear by it.
 
Thanks in advance for any insight.

The main practical use of oversampling is to move the Nyquist frequency so high that an anti-aliasing filter will cut off the part of the spectrum that might cause aliasing when the filtered signal is re-sampled to a reasonable frequency. Where are your plugins getting the signal to oversample in the first place? If it is an audio signal that has already been through the process in AD conversion there would seem to be little point. If it is a machine-generated/calculated signal that uses algorithms that put out a lot of ultrasonics then it may be necessary to clean them up, but why wouldn't the designers just build a reasonable amount of oversampling into the output stage? Letting the user choose from a range seems like a gimmick. 

This is an option on a number of Voxengo plugins I have. Yes, it is after A/D conversion.

It wouldn't really be cumulative.
 
If a DSP process creates/adds frequencies higher than 1/2 the frequency the processing is performed at, the stuff that's greater than 1/2 the sampling frequency will alias down into the audible range and sound nasty.
 
A given process only needs to be oversampled to the point at which those high frequencies are low enough in level that they will be inaudible. Given a reasonable margin for error, more than that is not any "better" in the real world, because once something is already inaudible, it's inaudible.
 
However people who have had their minds turned to mush by reading too much audio marketing material will argue that if it were audible then it would be better, so therefore it's "better", even if it's not, you know, audible.

The other main effect of "oversampling" (if I'm understanding your question right) is moving the decimal point for rounding errors.  This is one of the reasons oversampling sounds smoother etc. to some people.  Whether you or your audience can hear it on an average system is something you need to do for yourself.  Nowadays, with computers getting faster and faster, everything should be oversampled since computers no longer choke on the math.  It ought to be standard, rather than switching to oversampling, like Dimpro and Rapture.
 
@

"Why is it better?"
"Because it goes up to 11."
"But the knob's in the same position, isn't that the same as going up to 10?"
"No, it's better...it goes up to 11."
 
If available, I'll engage oversampling. If there are multiple choices, I'll stop at the one that produces an audible improvement if the next higher option doesn't sound better. If oversampling is not possible and recording at a higher sample rate sounds better, I'll do that.
 
 
 
 

AT
The other main effect of "oversampling" (if I'm understanding your question right) is moving the decimal point for rounding errors. 

 
It should be clarified that "oversampling" has nothing to do with rounding errors. Changing the bit depth is what affects rounding errors.
 
And like with oversampling, once the errors are inaudible (plus a reasonable margin), increasing the bit depth further doesn't make things any "better".

drewfx1
Once the errors are inaudible (plus a reasonable margin), increasing the bit depth further doesn't make things any "better".

But it goes up to 11. It has to be better.