Finally, I dare to write something about it, because previous attempts to be a bit triumphant about my approach to problems were always overtaken by reality...

The reason was my problem with rtcwake on the home server. I started looking for a cheap alternative to my motherboard, hoping that my rtcwake problems were only caused by the motherboard. What else could it be? Not the power supply, not the computer case, not the HDDs, because they ran fine under ZFS after resilvering. So looking for another MB. And since I still had some unused DDR3-DIMMs lying around, I specifically looked for a DDR3-MB. It ended up being a Biostar A68N-2100K mini ITX motherboard. Slightly less processing power than the previous board, but with DDR3 instead of DDR2 and moreover double the RAM: 8 instead of 4 GB. In short, a very suitable board to run Ubuntu Server 20:04. The only real pain point was that there are only 2 instead of 4 SATA connectors, and that was difficult because of my RAID with 3 HDDs. So I also bought a PCI-e card that gives you 2 extra SATA ports on your motherboard. Ubuntu must have the drivers in-house for this, and extra drivers were included for Windows.

Let's get to work then... put the server down on the table, and first placed the PCI-e card on the old MB, so that I could connect an extra HDD to copy some important files. Because the server was on the table, it no longer had a network connection, so I couldn't copy the files in any other way. At the same time, it was a good test for the PCI-e card.

Then removed the old MB and installed the Biostar, with the PCI-e card in it, and installed Ubuntu server 20:04 on it. The result was simply not good. Ubuntu runs fine on this motherboard, but the PCI-e card certainly did not and generated all kinds of error messages. That meant that this setup was not suitable for my home server!

In the meantime, it had become Wednesday and that was important because (until recently) something important happened on Wednesday: a backup of my home server on the old NAS was started. Because the home server was now downstairs on the table, without a network connection, I received a message on my desktop (as previously reported that it is now a system based on a Chuwi MB) that the NAS was a bit upset. And that got me thinking...

That NAS is a Synology DS411J, purchased in 2011, shortly after we were evicted from our home after a fire. To replace my then technically written off desktop after smoke damage, I bought a NAS with the advance of the insurance money, so that I still had "something". At the time, all HDDs were quickly copied to that NAS, so that few files were lost. The desktop and the then laptop were then removed up by the company that came to clean our burnt-out house. That NAS has been our main computer from 2011 to 2015, with 6.5 TB (actually 6.1 TB) of disk space and it ran pretty well. It was a very compact and quite cheap solution. But it also had limitations. The processor power is not really modern and the memory (only 128 MB DDR2) cannot be expanded. The amount of memory in particular means that the device is too slow to function as a nice, modern server, while you can install apps for DLNA and Downloads, but then really can't use it for anything. Purely for file storage it was a pretty nice server, if you didn't have too strong expectations. As soon as there was some financial room, from my mother's estate, I built a server myself, my home server, with a very nice MB with no less than 12 SATA ports and 16 GB of DDR3 DIMMs, I have written before about that MB.

Some things got me thinking. Actually, a good NAS is not that bad... it offers possibilities that are difficult to copy with a self-build server. On a NAS you have no problem with rtcwake and you can also define all kinds of RAID configurations. The standard NASses do not (yet) do that with ZFS, but the software is so good that they can also do something very nice with regular ext4 HDDs. They also regulate temperature management automatically. And, certainly not unimportant.... they constantly monitor the quality of the HDDs. Bad spots are neatly marked and not used further. This extends the life of an HDD considerably.
In short, on closer inspection, a good NAS is a good solution.
Only ... a good NAS also costs good money. And if you don't have that, you have to keep thinking. After much deliberation, the situation has now changed in such a way that my old NAS, the DS411J is now my main storage server. It now contains the 3 6TB HDDs that used to be my ZFS-RAID. In the NAS they run on a RAID-5-like setup, so with one disk security. But this NAS actually only talks to my home server, so it is purely for file storage. The home server acts as a DLNA server, download server, and additional storage for the desktop. What used to be the D: and E: disks on my laptop are now placed on the home server, along with the C: disk and also the I: disk (for all installation-type files). These dummy disks are backed up to the NAS daily with RSync.

This home server (with currently the MB that has been in it recently, and therefore only 4 GB of DDR2 RAM) switches itself off at night, with rtcwake. That now works reasonably well, via S4, so suspended on HDD. I still don't get how you can determine which rtc modes work or not, but I did find out how to discover what should work:

dmesg | grep rtc

returns the rtc modes that should work on a given system. And indeed!
That does not mean that all problems have been solved, they have been moved forward. S4 is a suspend mode, so will try to restore the pre-suspend state at the programmed time. If that fails, it will go wrong, which means that the system will simply remain in sleep mode and must be started by pressing the power button. That's not such a big deal, because it doesn't get stuck in that procedure either, and just starts up. I've been trying for several weeks now and have been able to reduce the times the system doesn't wake up on its own to once a week. I just need to make sure that when the home server wakes up from the rtc, it can access all the things it could access when the suspension started. In other words: the desktop must have closed the links to the home server in time.

We now have a stable, well-functioning setup and we have given the old NAS a meaningful old age. Once we really have the money, we will buy a modern NAS, and then the home server can also retire.

And as implied in the above... The home server is still running on the old MB (a Point of View (PoV) IPX7A-ION 330). That was the fastest solution if we focus on installation speed. What will be the fastest in use remains to be seen. The PoV has 4 cores and a somewhat faster CPU than the Biostar (2 cores), but only has 4 instead of 8 GB of RAM and has the slower DDR2 instead of DDR3 memory). The point is that everything is working fine. The Biostar is back in its packaging. After that, I will consider at my leisure what the final choice should be. Because of the hunch to use the NAS more intensively, it seems that the Biostar has been bought for no reason, but that is how it sometimes goes.