It is very interesting topic indeed.
Personally I tend to buy highest res that is available. Just because it is available 
with some exceptions… I explain those later. Higher resolution is not always better because I have encountered new hi-res loudness war remasters which are clearly worse than their CD quality originals, same is valid for CD quality remasters.
I have thought a lot about it and made a little bit of research also. Since beginning of digital audio CD quality - 44,1 kHz / 16 bits is considered gold standard. Some places note that 16 bit word length was proposed for studio use and for consumers 12 to 14 bit was considered sufficient. Converted to dynamic range 16 bits gives 96dB and 14 gives 84dB. Both these numbers are better than consumer grade analog formats, like tape or vinyl, and are pretty good for studio work. Now-days it is very common to use gate or noise reduction to create perfect black background, it wan’t very common back then. Also as many know, for computers 16 bits are much simpler than 12 or 14 bits. 12 and 14 bit words just waste space, because padding bits are needed. So 16 bit is adequate.
Basically same thing goes with sampling rate. 44,1 kHz gives us 22,05 kHz bandwidth. Plenty of room and again better than consumer grade analog formats.
All that I wrote above is for PCM format. Also it must be understood that these numbers are probably also based what technically possible back then. In general, whole system was at its limits and had many now known and understood problems.
When hi-res files arrived I started to look more into subject and discovered very interesting fact from some medical non-audio book (I have to dig link out). It was something like this: if signal must be precisely sampled, then sampling rate must be at least 10 times higher. I started thinking how, why and how it can be connected with audio. And quite quickly realized how and why. If we want to sample something and signal shape is not very important, then nyquist theorem works well. On the other hand if we need to sample something very precisely to examine waveform shape(heart beats, seismic waves…), then we need to use at least 10 times higher sample rate. Simply put: there is difference between rough and precision measurements. And what is digital audio? Just measurements. So CD-quality is rough and hi-res is precision measurement. I asked about this from people who didn’t deal with audio, but did general electronics design. They confirmed that this is in fact true and well known fact what was used when signals were analyzed or measurement and test equipment were chosen. After all that I did visual comparison between hi-res and down sampled waveform and indeed hi-res waveform contains visually more details at higher frequencies. If this theory is considered, 44,1 kHz is sufficiently precise only up to 4,41 kHz - is it low? Is it high? Is it enough?
When digital audio developed further and digital mixers, recorders, processing hardware and software came available suddenly higher sample rates were needed. Only then was discovered that it wasn’t possible to make DSP tools based on CD quality standards. To make these DSP things work, internal up-sampling was needed. It was probably at that time when somebody decided to work higher rate from beginning to end. Before material was internally up-sampled, then processed and then down-sampled. Working at higher rates from beginning may simplify whole system and as we know it simplifies digital filters in ADC and DAC parts. In general - hi-res takes of strain from digital chain. Wind instruments and percussion can create bursts of almost white noise and alias filters must be very steep. At higher sample rates there is more room to play.
It is logical that in development, scientists try to refine processes to more precisely capture and reproduce signals, also audio signals. We can argue, does our ears can hear differences between CD-quality and hi-res audio or not. More important are hi-res audio technical benefits and advantages for those who design ADC and DAC chips, also digital audio processing equipment. As I mentioned earlier, alias filters were headache back then. I have heard that it is simpler to design slow roll-off filter than brick-wall filter
For conclusion, I personally like hi-res if done correctly. I don’t know, maybe I hear difference between CD and hi-res when listening percussion or wind instruments, but it is when I listen with electrostatic STAX headphones.
PS: Now about that exception. There are a lot of fake and up-sampled hi-res that should be avoided. Up-sampling don’t bring nothing back - what is lost, is lost. It just costs more to buy up-sampled file. I have encountered this up-sampling thing mostly in DSD world, where DSD1024 is big thing… do your research before buy, maybe it is possible to save a bit
Sorry about my bad english.
