> If you're dealing with cassettes or vinyl or reel to reel, and you need to do any time based corrections (stretching/shortening), higher rates will definitely give better results.
EmRR, there's no necessary reason why this should be so. If that's your experience, I would think that it's due to the particular software that you're using and not a general rule.
As long as all the energy of a continuous (analog) signal is below 1/2 the sampling rate, the complete information of that signal is contained in the samples, subject only to the quality of the conversion in terms of noise, distortion, etc. Any representation of that same signal at a higher sampling rate would contain no further, actual information about the signal. More numbers, yes--but not more information about the signal itself, just a more verbose restatement of the same information. As if I kept making the same point in various ways in the messages that I post, or as if I wrote in a redundant fashion--you get the idea.
One of my favorite processing devices is a Cedar AZX+ azimuth corrector. It has only digital inputs and outputs. You feed it a stereo signal at 44.1 or 48 kHz, and it resolves the time relationship between the two channels down to 1/100 of a sampling interval. According to the usual audiophile misunderstanding of digital audio this device cannot possibly exist, because supposedly a sampled waveform contains information "only at the moments when the samples themselves were taken"--yet here it is two feet to my left, and it works very nicely.
--best regards