The last time sampling rates were hashed out on here, I posted a little test I did with an instrument part exported at 48kHz and then again at 96. It was a patch from an AAS sound bank. Anyway, it really did sound inferior at 96. But I've also tried whacking some Reaktor synths up to 96 in the Reaktor settings and they sounded noticeably fuller and creamier. I've also had a Z3TA+2 patch sound worse on 2x over sampling. So I guess everything needs a case by case judgment.

So unless you are using plugins that have taken shortcuts and neglected to include oversampling in their code, then converting an entire audio session to a higher rate would make your mix take up more processing power without adding any sonic benefit.
 
That's from the link James posted #136. Very good read.  Just re conferming what I already understood about the topic.
And now I'll return to my work which has either been 44.1 or sometimes 48. I like 48 because my Sony DAT players used it and I still find those are the best masters I ever made.
There did seem to be an audible difference back then. But with my new equipment, I just can's hear it now. So I mostly work at 44.1 because it works best in my system and I personally have never seen the point of going way higher.  48 yes, 60 possibly, 88.2, Oh what  the heck,, but 96 and over ,,, nope) . After reading dozens of articles on the subject, including the one above I see why I have made my choice, and made it wisely to suit MY needs. 
 
I thought this little snippet from the article was worth a gander. 
 
192kHz digital music files offer no benefits. They’re not quite neutral either; practical fidelity is slightly worse. The ultrasonics are a liability during playback.
This runs counter to many initial intuitions regarding super-sonic sampling rates – my own included. But the evidence is there. Since analog circuits are almost never linear at super-high frequencies, they can and will introduce a special type of distortion called intermodulation distortion.
This means that two super-sonic frequencies that cannot be heard, say 22 kHz and 32 kHz, can create an intermodulation distortion down in the audible range, in this case at the “difference frequency” of 10kHz. This is a real danger whenever super-sonic frequencies are not filtered out.
 
 
 

drewfx1
It means that under at least some conditions Z3ta+ can generate frequencies > 24kHz (i.e. 48kHz/2), which causes aliasing (imaging) distortion because it is > 1/2 the sample rate. If you run the same code at 96kHz, only frequencies > 48kHz (= 1/2 * 96kHz) would cause this type of distortion.
 
My recollection is that Lavry was talking about non-oversampling converters being less accurate at higher clock rates (due to analog components accuracy in measuring declining at higher rates), and balancing this against the difficulty of creating steep purely analog filters. This is a converter design issue and the game changes when we are talking about oversampling converters. And once the signal is digital, the analog limitations are irrelevant.
 
 
Unfortunately it is more complicated than might be ideal, but it may help to divide things between the converters (ADC/DAC) and DSP processing.
 
For converters, higher sampling rate = higher frequencies can be present in the signal, and that's pretty much it. So there is no benefit to increasing the sampling rate > twice the highest frequency needed (plus an appropriate margin for error).
 
But any type of DSP that creates frequencies > 1/2 the sampling rate that the processing is done at will create (aliasing/imaging) distortion, so increasing the sampling rate can improve the quality. To complicate it more, note that this doesn't apply to every type of DSP, but only certain types. Ideally, the programmers would take care of this kind of thing behind the scenes (and often they do), but there are cases where they haven't.
 

 
Is there any way of comprehensively determining the highest frequencies that each of the programs included in Sonar can generate, so that we can see just how prevalent these circumstances are (at least within the Sonar package), and/or which of said programs oversample behind the scenes, and which (if any) do not?
 
Might the Bakers be able to assist here?...

johnnyV
This means that two super-sonic frequencies that cannot be heard, say 22 kHz and 32 kHz, can create an intermodulation distortion down in the audible range, in this case at the “difference frequency” of 10kHz. This is a real danger whenever super-sonic frequencies are not filtered out.

 
That's like a beat frequency issue isn't it?  The human ear can do similar things, even if each ear is hearing a different tone, so the beat is actually generated in the brain (http://en.wikipedia.org/wiki/Binaural_beats).
 
Humans even 'fill in' a missing fundamental frequency, e.g. if 2f, 3f, 4f...is presented to the ear, we'll perceive the pitch as f (http://en.wikipedia.org/wiki/Missing_fundamental).

sharke
The last time sampling rates were hashed out on here, I posted a little test I did with an instrument part exported at 48kHz and then again at 96. It was a patch from an AAS sound bank. Anyway, it really did sound inferior at 96. But I've also tried whacking some Reaktor synths up to 96 in the Reaktor settings and they sounded noticeably fuller and creamier. I've also had a Z3TA+2 patch sound worse on 2x over sampling. So I guess everything needs a case by case judgment.

I think it's important to differentiate between something sounding "worse" and "more accurate." When I posted the example of a non-oversampled z3ta+ 2 recorded at 96 kHz vs. 44.1 kHz, I noted that the one recorded at the lower sample rate was more musically useful because the one at the higher sample rate was excessively bright. However, the excessively bright one was a more accurate representation of the sound I had programmed.

I used to record at the highest rate my soundcard would allow, but after the novelty of the bigger number  delusuon wore off, i record most things at 48khz and sometimes 44. Maybe if i did more actual paying work that required heavier processing I would go bigger, but I strain to find even a slight perceptual difference to justify the extra space those files take up.

dubdisciple
I used to record at the highest rate my soundcard would allow, but after the novelty of the bigger number  delusuon wore off, i record most things at 48khz and sometimes 44. Maybe if i did more actual paying work that required heavier processing I would go bigger, but I strain to find even a slight perceptual difference to justify the extra space those files take up.

Remember that you can record at 44.1 kHz, but transfer your synth and sim-related parts to 96 kHz, then render back down to 44.1 kHz and insert in your original project. Same taste, less filling 

Many beat frequencies of inaudibly high frequencies are audible. If you are recording a live orchestra, you'll capture those beats, so you're OK. Many of us record instruments separately, and if we don't capture the inaudible frequencies, the audible beat frequencies are never generated; never heard.
 
I've never heard a recording of brushes on cymbals that sounds convincingly like live brushes on cymbals.
 
It probably doesn't matter, though. We've been conditioned to listening to music that has been mangled by mics, pre-amps, amps, speakers, and venues. We very seldom hear music right from the instruments that make it, and even those instruments are optimized to emit audible frequencies. It may be like even temperament. It's wrong, but we're used to it. I even LIKE it. It is very nice to hear a Barbershop quartet hit a natural scale dominant 7th, but I've come to rely on that slightly ooky feeling I get from an even tempered dominant 7th in my compositions.
 
To be responsive to the question, I use 96/24 for live vocal groups, horn groups, string groups and some synths; 48/24 for amped groups and single recordings that will be mixed to form the group. I may just go to default 96/24, since it doesn't seem to exact any great cost with my equipment and technique. I do remember the fiasco in which I climbed a large learning and equipment curve to go 7.1 192/32, only to find zero market for the product at my level. As has been said, if your audience is 128 mp3, just relax. ;)
 
I apologize for the use of the word "ooky" in this post...damn...twice now.

After reading these, I decided that for the money I'm willing to spend on converters, 48K is plenty good for me.  The first references the second.
 
http://www.trustmeimascientist.com/2013/02/04/the-science-of-sample-rates-when-higher-is-better-and-when-it-isnt/
 
http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf
 

JohnDubats
Many beat frequencies of inaudibly high frequencies are audible. If you are recording a live orchestra, you'll capture those beats, so you're OK. Many of us record instruments separately, and if we don't capture the inaudible frequencies, the audible beat frequencies are never generated; never heard.

This is commonly stated, but it's not true.