I can only speculate that maybe the circuits in the Ethernet receiver react differently to slower edge-rates and poor signal integrity. Maybe the setup timing margins are smaller. This might cause the propagated signals to have more jitter.To add to Steve's comment, it seems very conceivable to me that in some and perhaps many less than perfect designs a small but potentially significant fraction of the signal energy received by the Ethernet interface may find its way **around** that circuitry (via grounds, power supplies, parasitic capacitances, etc.), and end up introducing noise onto the signals which control timing of D/A conversion.
All of the energy of a signal doesn't necessarily follow only its intended/ideal pathway. Especially when the signal contains spectral components at very high frequencies, as in the case of Ethernet.
Regards,
-- Al