Roon's update sounds really good

does the buffering take care of the interruptions like matching the clocks?
Yes.  Ideally, you have more data than is asked for.  Say 1 seconds worth.  While the input/output must be balanced, on average, it does not have to be exactly right all the time.  So, the DAC may continue to feed off the buffer even if the incoming data is randomly delayed by 0.01 or more seconds.

Oh, lastly, Ethernet uses balanced, ground isolated connections by design. Ground loop issues are a real problem in offices and labs, given the long runs possible so the whole Ethernet chain is designed with this in mind.

However, routers and switches usually have noisy power supplies which can contaminate your AC line.

this is this is a quote from the owner of Auralic:

“The network router, the NAS and other network device are directly attached to the router using Ethernet cable which can be quite noisy. If the streamer is attached to the router using Ethernet cable then the noise can easily be transmitted to the streamer over this physical link. The regular CE, FCC and other EMC standards only guarantee the device will work stably, but when we are talking about sound quality, we are actually talking about EMI noise that is 1/1000 of the EMC’s standard, so it affects [sound quality] a lot. That’s why audiophiles say different NAS or different routers sound different.

The Ethernet cable is [also] a big problem. It runs long and can pick up low-frequency noise which cannot be filtered out. This noise will introduce jitter at low frequencies which no PLL circuit can get rid of and it will affect sound quality a lot.

WiFi will sound better if:

The engineer that designed the streamer knows how to design the WiFi module right. ‘Right’ means the module does not introduce noise to other audio circuits sharing the same ground circuit;

The WiFi signal between streamer and router is strong and stable. This is because if you have a poor WiFi signal, the WiFi protocol will try to fetch the data again and again due to large error rate (Yes, we use TCP to transfer music data which has error correction). If there are too many retries, it will add additional load to the processor and potentially generate more EMI noise.

The reason that many audiophile companies claim Ethernet to sound better is that they are not capable of designing a good WiFi circuit and they have no such background knowledge (poor WiFi network). All our streamers use existing Intel WiFi modules because they work well on a Linux system and they are well designed with minimum EMI noise generation and, most importantly, we know how to design the circuit around it correctly.”

There's no way to get a network receiver in a device that isn't noisy. It happens. However, I agree that the cable itself can be a source of RFI/EMI interference, so it's good to route it away from analog interconnects and power cables.

I would imagine however that Ethernet, lacking an actual RF tranciever, would be easier to design a low noise interface to, but I've never had to do it.  I would take this information into consideration, but not without some skepticism. :)

The Ethernet cable is [also] a big problem. It runs long and can pick up low-frequency noise which cannot be filtered out.

It’s balanced, twisted, shielded, ground isolated. If it wasn’t perfect most offices wouldn’t function.