Well, that's software developer-led cowboy coding for you: prioritization based on weekend impulses rather than user-driven urgency. Also, the hobbyist and volunteer excuses don't fly because a professional effort with a saleprice of $0 has to compete with non-free and is subject to forking and loss of interest if it founders or stalls. Also, I don't think there are enough NetBSD devs. It's main users are some ISPs for network infrastructure IIRC.
BTW, FreeBSD doesn't even run usably on RPi4, it boots and that's about it. RPi3 support is somewhat better, but still worse than RPi2.
I hope all FOSS OSes, including Linux, get and stay organized and productive, but not heading toward a greater unbalanced power law distribution (pun unavoidable) monoculture where only Linux dominates, because then we're in real trouble.
Opinion: which is why I think the limited volunteer hours available for *BSD projects shouldn't be fragmented across NetBSD/FreeBSD/OpenBSD/DragonFly BSD.
This makes NetBSD a good choice if you intend to have fun developing an OS, but not a very good one if you want to make a living running software on said OS.
It's interesting, because if you look at OpenBSD, which forked from NetBSD in the '90s, you see a similarly spartan feature selection that can seem from the outside to be selected on developer whim and sometimes with priorities that don't match everyone who might want to use the system. However, rather than letting the features rot, they often just remove them. So for example, some years back they just went and removed Linux binary emulation; no doubt it was tremendously useful to some people but they weren't interested in maintaining it, so poof, one release it was just gone, and if you weren't paying close attention to the platform this news might have just come as a shock one release. In some ways that's more honest but if you happen to rely on one of these features it can feel destabilizing.
I write this not as a criticism per se; I have generally liked every *BSD system I have used.
I'd just like to add that Qemu can use NVMM (-accel nvmm) in which case almost all of Qemu's existing documentation is still relevant. You do still have to build it yourself however since it's a WIP package[1]
Same with ZFS, I've not found it to be any different than the ZFS implementation I was using on SmartOS previously. I'm sure there are some differences lurking as it's not the latest version of ZFS but on the surface it's behaved as expected.
These are not excuses, NetBSD does not have a lot of resources, documentation could definitely be better. I'm just really excited about this release as ZFS and a type 2 hypervisor have been the 2 things I missed most in other operating systems and both have been working well for me.
On NetBSD.org there is a link under documentation called: The guided. Neither ZFS nor NVMM seems to be documented. ZFS does have manual pages, but if you want to use NVMM you seem to be on your own.
There's a ton of interesting things in NetBSD, but you don't get the feeling that there's a coherent plan behind it all, like you do with both FreeBSD and OpenBSD.