I did some initial testing.
I got a 69C reading on the MB sensor with the CPU long term power limit (PL1) capped to just 20W after 50-60 minutes of Linux kernel compiling + playing a 1080p movie with mpv with some demanding post-processing for additional GPU shader workout + light browsing.
The config is: 9600K undervolted with -90mV offset, LP53 + Noctua 9x14, Asrock Z370 Fatal1ty, 32GB DDR4-3600, Samsung 970 Pro 1TB in the x16 slot, HDPLEX 200W DC-ATX, Noctua 4x10 on one side of the case.
So I guess if the CPU is putting out 20-25W heat the whole system is around 35-40W when heavily loaded.
There's an important caveat: the intended use case is indefinitely long running full loads on all cores while staying very quiet. So this how I got the 20W power limit.
I don't like the noise coming form the Noctua 9x14 at 40% RPM and above so the CPU fan is capped at 33% RPM no matter how high the CPU temp is. Similarly the small 4x10 case fan is capped at 33%. The fans go to 33% above 90C CPU temp. With these the noise is alright with heavy loads, might be better than most systems, and the system is inaudible when doing light tasks, with the fan at 25% or 800 RPM or so at minimum.
So there must be very little airflow in this already airflow limited case. When idling for a long time, the MB sensor is around 48-50C and the CPU is 55-60C. With heavy load the CPU can boost to 4Ghz an all cores for a short while when starting from a lowish temp, then the PL1 kicks in, it goes to around an average of 3.5GHz for a few minutes, then it goes further down and settles at an average 2.5GHz long term.
The CPU temp then stays mostly in the 88-97C range, with thermal throttling typically not happening. The MB sensor slowly rises towards 70C over time. Only tested for an hour so far though. The SSD without heatsink is reaching 100C on one sensor. The top lid of the case is above 55C, just like the temperature of air behind the rear vents when sticking inside a generic multimeter temperature probe. This is still with 20W long term CPU limit. When idling again the cooling off is very slow, like it needs up to 5 minutes of idling to shed 2C. The CPU goes below 90C quickly and with that the fans go back to minimum RPM which doesn't help with that.
Will it settle at 70C on the MB sensor when running for many hours is still a question. Otherwise what should be the TDP limit for the CPU then. We're almost in fanless NUC case territory here. Another thing, when running so hot, after a while you can smell that the system is running hot, which is unfortunately not very nice, there's probably some outgassing from some components at those temps.