What is the added value of a streamer over a networked dedicated Computer

I have direct experience in running dedicated Mini as both server/streamer and server only. Mine I7, PCIe ssd, Uptone MMK and JS2 LPS, running Roon as server only.

Based on your description greatest liability of present Mini is doing usb rendering within Mini, Mini very weak as renderer. You need to create second ethernet port via bridging of Thunderbolt. This https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/ will explain how to do this and also provide great info as to streaming in general! Believe me,  the bridge mod alone will totally transform sound quality!

So, what above means is Mini as currently implemented is far from optimal, the second ethernet port gives you ethernet out to a streamer which will do the critical function of usb rendering. Yes, you could do without the bridge and simply use a switch or router and then go from that to streamer, but this not nearly as good as bridging. This also comes from personal experience via use of audiophile switch.

Bottom line, using Mini for streaming not good, inferior renderer, relatively noisy device regardless of any mods one may undertake.

And then you have liabilities of Mac OS, from what I understand Audirvana Plus does a modicum of OS optimization on it's own, but much more can be done, need third party app wipe, go into Terminal to disable other processes such as System Integrity Protection (SIP), etc.

So, assuming you're willing and able to do all the above, you can have pretty nice server at this point, but a separate streamer still provides very large upgrade for sound quality. But then Audirvana really built for one computer setup, aka server/streamer combo, the server does the rendering. Roon optimized for two computer setup, the Mini or some other server and separate streamer. IMO, running Mini as one computer setup with Audirvana is far from optimal since requires Mini to do rendering.

So, you have the above considerations if keeping Mini. Another path would be to get rid of Mini altogether, if keeping Audirvana get server with optimized usb rendering, assuming you're continuing with usb. I'd seriously consider rendering capabilities of any server you consider for purchase, rendering is second most important component of streaming chain after dac. Point I'm making, many off the shelf servers don't have optimized rendering, which then requires adding a streamer or band aid usb filter 

 I should also add you could continue to use Audirvana with your Mini and streamer. My present setup running Roon utilizes Mini as server only, use Sonore OpticalRendu as streamer, extremely high quality streaming sound quality, beats my pretty nice analog setup. You could use OpticalRendu in your present setup, OpticalRendu has many choices for endpoint implementation which includes ability to use Audirvana. Also possible to implement other streamers with your present setup, just have to do the Mini mods I referred to above for optimal sound quality.

While Audirvana built for one computer setup, you'll still get benefit of superior rendering via separate streamer, problem is you're adding needless complexity. If wed to Audirvana I'd get very nice one computer setup vs keeping present Mini and adding streamer. Roon would be more advantaged with the added streamer. In either case, your Mini needs further mods for optimal sound.

In the interest of clarity, since many call what I consider servers, streamers. What I'm speaking of is a DEDICATED streamer, many of us know this, but this is what is used in two computer setup. Only the DEDICATED streamer and/or really high end server/streamers do optimized rendering.

To be clear, the server provides music player library and music player processing ONLY, the DEDICATED streamer does only the rendering.

Relatively easy to understand optimal input on dacs, not so easy with the computers serving dac. First off, servers and streamers both computers, and it is the rendering service they provide that makes all the difference in obtaining maximum potential from dacs and streaming chains. Servers or one computer setups may or may not provide optimized rendering, many run rendering ports directly off motherboards, not good since these motherboards must run many processes allied to music players and OS, the more processes run the more noise produced, not good. The streamer has very little to do in comparison to server, motherboard runs very elemental OS, with focus on rendering, which means nice power supply and good clock. Servers very rarely provide this level of rendering, the few that do are easily $10k plus.

Get a nice streamer, server becomes much less important, not saying still not important, only less. In order of importance, dac, rendering, server, network.

There is one other way around two computer setup, streaming dac. Assuming quality rendering within dac one only needs to concern oneself with server.

I'd agree with those stating general service computer doesn't belong in quality streaming setup. There are means and paths to go diy through self builds and modifications of off the shelf GS computers, but one needs computer savvy. Add up time spent gaining knowledge, actually performing mods, expertise to carry out those mods, parts needed for these mods, Nope, vast majority better off getting off the shelf servers/streamers.