Search the forum for more assertions about the audability of using higher than 44.1ksps.
However, know that many of your favourite plugins may already be doing incoming upscaling and outgoing downscaling because they are internally running at 192ksps.
Bob Katz, author of Mastering Audio, recommends recording or upscaling to 96ksps and downscaling only at final rendering, because good filters are hard to come by in most prosumer equipment. He also cites research showing that digital processing artifacts are more outside the audio passband with higher sample rates, and thus easier to filter out.
But, the basic presumption about the inaudability of differences between 44.1ksps and higher rates is based on one conversion with no complicated processing. Most projects will have multiple FX, etc, with thousands, if not millions of calculations, each possibly producing artifacts. The real test is to do the same project twice, with one completely done at 44.1ksps and the other completely at the higher rate, with only a single downsample at the end.
There is also another presumption that up and downscaling is really bad, but with no research cited to support that.
Also, as p j geerlings posted in a thread last year, phase relationships are not going to be accurately reproduced using a sample frequency twice that of the highest required frequency. That is, using a sample rate of twice a frequency will only tell that that frequency exists but CANNOT resolve its phase. Several sample points within the cycle will still only crudely resolve phase. This last is possibly the most significant for using multiple FX, EQ, etc, as the sound will be different depending upon the number of samples used for calculations. A 192ksps rate will still only resolve phase of a 10kHz tone to 18 degrees. What will that do to a 1kHz square wave, and then think of what 44.1ksps will do. What square wave?
It is obvious that FX that are trying to precisely emulate specific analog gear are not going to be phase accurate in the higher frequencies using only 44.1ksps, which is why they probably upsample before the thousands of internal calculations and downsample at the end. Even than, there will be some degradation of quality.
Again, I reiterate, do not quote a simple, single conversion up, zero intermediate processing, and single conversion down as proof that a complete project does not need to be done at a higher frequency than what the finished product will be. Just because someone has taken the trouble to set up equipment and measure one simple scenario does not mean the results apply to ALL scenarios. All that those A/B experiements show is that it may not be necessary to LISTEN at greater than 44.1ksps, but says NOTHING about whether the thousands of processing steps to generate what is being listened to should also be completely done at that rate.
In the end though, people will believe whatever they want to rationalise their biases, and proof and rational thought be damned.