Oh god, when I wrote that I knew somebody would get pedantic about it. Yes, it’s the speed of light “in the medium”, aka the “wave propagation speed”, aka 1/sqrt(epsilon*mu) and is relative to permittivity of the medium.
Since we’re talking Ethernet versus analogue here, we’re firmly in the realm of “copper” - nobody’s using fiber for this stuff except the MADI folk, as far as I’m aware (though there ARE other fibre-based audio transports - I’m talking specifically common-use live-sound tech and focused mostly on Ethernet-utilizing tech here anyway).
Anyways, the point remains: the latency of AVB / AES50 / AES68 / Dante / etc. is the sum of the A/D latency, the transport buffering, the wire speed of the signal, the receive end buffering, the internal summing/routing, etc… And the latency of analogue mixing is simply the wire speed of the signal through the chain (yes, if we want to be super crazy we can count phase shift through the EQs and other stuff, but that exists on the digital level too). Therefore it’s impossible that, compared to analogue mixing, any digital (copper-based) transport can have lower latency, it could be theoretically equal, but never lower.