You cannot (except by switching manually), because that doesn’t make any sense.
I have bunch of files at 120 kHz sampling rate. There is also 32 kHz-based standard sample rate family with 32/64/128 kHz sampling rate. I have also a DAC running at 40 MHz sampling rate.
There is no reason to be constrained to few sampling rates. Best would be to optimize sampling rate and word length for each song specifically based on spectrum analysis of the content. I’ve done some of such and it works great for saving some disk space.
There is no need to isolate HQPlayer’s NAA traffic from other traffic. I have everything as single network (apart from couple of special purpose VLAN’s but those of course go on the same wires). Total six switches and four wireless access points. Having HQPlayer running on a laptop and playing 8 channel DSD256 over WiFi to a wired NAA works perfectly fine without any difference on sound quality.
If you have enough internet bandwidth and create a VPN, you could put NAA traffic travel over VPN tunnel to other side of the globe and still the sound quality would be the same. This is no different than playing something for example from a Tidal server (in such case, you have the stream mixed with lots of other people’s internet traffic).