7nm would be very much set this year. Last year would have been 14nm, and that's why I crossed that one out. I agree that doing it in 14nm would not be feasible.
I made a test built on Alternate with similar powerful hardware as is expected for Scarlett and got to a price of around 800€ - and that's consumer price. Microsoft could certainly shave off at least 100€ off of that, if not 200€. At 600€, they could sell it at $499 for the first year(s) until the prices drop for them. That was the modus operandi for gen 5-7, and the losses ain't too big to not recover them with the software sales.
Dreamcast needed about twice what it was selling for, and we're far away from that. Plus, Dreamcast got plugged because Sega was bleeding money everywhere, not just on the console. There's no chance that could happen with Microsoft anytime soon.
Why wouldn't the next gen XB console come out in 2019 then if it would end up more like the 360 era? Zen or Zen 2, Navi, PCIe 4.0, cheaper SSD's, etc, are all going to be on the consumer market in the next couple of months. Why would MS wait until holiday 2020 if they could have launched late 2019? Maybe the 7nm yields still aren't high enough yet and the costs of losses would be too much this soon? Maybe they are waiting for 7nm+ on euv and there is a logical reason why they are doing so? Maybe TSMC just doesn't have the capacity or is booked solid this year? Why would they use Zen and not Polaris or Vega, when Zen 2 and Navi are about to launch? Maybe because they would have had to pay a lot to get Zen on 7nm and Navi moved up so it was ready in time? Maybe they didn't want to pay that, or it happened as fast as possible anyway, but they also didn't want to try and have to use another semi custom Polaris or Vega? Maybe the costs are too high overall for everything right now and so they decided to wait a year? Maybe they feel the power and tech narrative is super important, and even if they couldn't best PS5, being as close as possible is important, considering what happened with XB1? Who's to say 2020 is the right time since that will only be 3 years since XB1X? How long is long enough, or too long between consoles? There's a tonne of reasons and more as to why launching in 2019 may not have made sense and would have been a bad idea, even with the unfortunate position XB1 would be in at this point without XB1X.
You're focusing too much on the costs, and assuming the best, when time line, performance, marketing, and brand image would be just as important if not more. While MS wouldn't suffer the same way Sega did due to their online infrastructure, in terms of hardware, it would likely end up a Dreamcast. There's even articles about Halo supposedly being proposed for PS4, and now Phil is talking Gears 5 potentially. Why do that if you're planning on selling enough of your own hardware?
The Canadian National Anthem According To Justin Trudeau