I have seen a few posts from people claiming the sound quality is different / inferior to some application or other without qualifying this in any way whatsoever. What I am most interested in is understand how and why that might be so.
If a user is using some kind of processing and (for example) another application has a better sounding EQ algorithm, better dither algorythms, better upsampling, then this would make sense. As an audio engineer, there are for example certain EQ plugin that I prefer the sound of. So, do you mean a better processed sound, or a better straight through sound (no EQ, upsampling, volume etc)?
If a better straight through sound, then one has to question exactly how the end DACs are being driven. Roon (https://community.roonlabs.com/t/raat-and-clock-ownership/6915/5) have described how they interact with and manage data clocking which is what I would assume to be the most likely source of subtle no-processing sound quality differences. From his description it seems to me they have taken the correct approach in effectively extending the data transfer behavior of USB over RAAT and thus Roon should not be introducing any jitter of its own making (due to forcing the PLL in a DAC to continually adjust, or by forcing async resampling) which both can introduce degredation.
If there is indeed a real sound quality issue, then those who are committed users of Roon may wish to consider passing on specific cases to Roon support to isolate specific cases of real differences.
Generic comments such as x sound better are not helpful. Comments such as X sound better when I set the same EQ settings on both are maybe starting points for further analysis if a specific track can be referenced. The ideal of course if there is an independent means of performing high quality capture of the resulting audio output to yield real analysable evidence, but few of us have such a capability (ie a high quality ADC that can operate at least 192K/24bit), but provided the information you do have may allow someone with the right equipment to reproduce a real issue at a specific point in a specific track if indeed there is an issue. If there is - then great - the right people have something specific to work with. If not, then the right people have some hard evidence to put some of these issues to bed. Win all round I think?