@MC - I don't think I have bat ears at all (although they certainly may be batty!!!); I didn't even offer my medical defect as a possibility until later on. When i offered this as part of the story, I did so with a dose of incredulity myself. So let's please agree that we should get this out of the picture.
This is why I offered the recording. By the way, even my friend could hear that recording. However, he cannot hear that when putting his ears near the tweeters, played back at the same level that I recorded it at. It is VERY soft.
Beyond that, my playback chain is very simple:
Two different streaming chains available; both start with a Synology NAS, and neither are using methods where the files are being "pushed" upstream, but rather the files are being "pulled" over a wired network connection, directly from the filesystem.
The first is: Synology NAS --> gigabit ethernet --> Auralic Altair used as a streamer, with all processing capabilities disabled --> USB --> RME ADI-2 DAC (new edition). I have tested this configuration using the RME Bit Tests, and all tests pass.
The second is: Synology NAS --> gigabit ethernet --> Windows 10 PC with Audirvana used as a streamer, with all processing capabilities disabled --> USB --> RME ADI-2 DAC (new edition). I have tested this configuration using the RME Bit Tests, and all tests pass.
I am using Balanced outputs from the RME, and for "Line Out" my settings are as follows: Source: Auto; Ref Level +7 dBu; Auto Ref Level OFF; Volume 0.0dB; Lock Volume ON. I have no EQ or loudness or any other effects enabled on the ADI-2. Oh, and as I mentioned earlier, DSD detection on or off doesn't have any meaningful impact on the results for me. The playback sounds pretty much the same either way.
In all cases with speakers, playback is through a McIntosh MA8900, to Salk SS 9.5 speakers. Headphone playback is to Sennheiser HD280 Pro connected directly to the RME ADI-2 standard headphone jack.
I have felt since starting this test that there was significance in the fact that all Bit Tests passed (and the Bit Test files are in the same subfolder as the Xiph files, going through the same playback chain), since that should prove that there is true file transfer integrity. I had actually thought that this was the most significant thing of all; hence why I left out the other details until now. @MC, if you can shed light on how the passing of the bit tests might be irrelevant in this scenario, that might shed some light on this for me.
Please remember that using either of these configurations, I can still hear the sounds you hear above in my headphones connected to the RME ADI-2, albeit at a lesser level than my speakers, from which I recorded the example. This is likely because my headphones lack high frequency response compared to my speakers. I am also able to hear artifacts on my speakers - at a notably reduced amplitude in comparison, as previously noted - using the same chains as above, but played back on my McIntosh DA2 module instead of the RME ADI-2 DAC.
I'd like to make one more point here - all of this is very difficult to hear in the real world. Even I could barely hear the artifacts while standing to make my earlier iPhone recording. The iPhone recording only serves to amplify the effect - and even I am surprised by how well the iPhone amplifies the effect! The point of my recording was to "exaggerate for clarity" of this conversation, as it were.
It appears that @DominicPerry is experiencing the same issues.
I hope this helps clarify things.