@Honken and
@Jinjuku, I think what we agree on is that any network-level transport used today is highly isolated from individual packet or bit jitter. It could be TCP, STCP, UDP, QUIC, etc., at the transport layer. Network streamers do not use real-time digital audio techniques and the packet payloads, as pointed out, could be highly encoded and processed from digital audio. This is why, I think, many (most?) of us in this thread are incredulous that clock and voltage regulation changes to an Ethernet switch will make any audible difference.
There are real-time layer-2 protocols, even some that work over enhanced Ethernet. There are some truly remarkably low-latency layer-2 link layers too used in high-performance compute (think super computers). But none of that is necessary for network streaming.
@Murmon has provided one concrete example of upgrading a Zyxel GS-108B v3 switch with a new clock and a new voltage regulator. I've looked at "newclassd" website, and cannot say I'm impressed with those parts. A +/-1 ppm clock is not cutting edge. There are 10 ppb (i.e. 2 orders of magnitude better) OXCO clocks available for less than that part from newclassd. And the vreg looks like a 75xx plus some filter caps. Anyway, there is a specific straw man put out. I'd suggest that if
@Murmon could send Amir that switch to compare against an unmodified one (which only cost $35 or so), we could see if there's anything audibly different.