Hello jmase. It is good to see you around.
I don't have answers, just opinions

What do Haiku OS, OpenBSD, Redox OS, ReactOS (and so on) have in common? They are opensource.
And they all have a active development community.
Just as FYI:
- HaikuOS (Microkernel)uses "NewOS kernel" developed by Travis Geiselbrecht. Travis worked on the BeOS kernel, Haiku OS Kernel and now he is working at Google in the FuchsiaOS project (Possible on the Zircon Kernel)
- ReactOS (Microkernel) has the "ntoskrnl" which is a kernel to be compatible with the NT Kernel.
- Redox OS (Microkernel) has a kernel made in rust.
- OpenBSD (monolithic)
I agree, if it is open source it has a better change to survive over time.
Opensource could make it interesting for people to look how it was made, get new people interested
in creating a modern (64bit) OpenOS/2.
You all know I'm a supporter of Open Source since a lot of time. Open source allowed us to keep using OS/2 today, we had took a lot of the Open Source community. It is not for me as "Open = Good" and "Close = Bad", it is more like "Open = Better", since it is better to have the source code and binaries instead of only the binaries.
Maybe a petition to IBM to opensource what they can would work now, so many years has passed...
(or ask them if the leaked sources can be used 
I am tempted to request it again to IBM, this time Arvind Krishna, the OS/2 sources, or partial sources, the OS/2 PowerPC source, permission for the leaked source or even a license change to the IBM device driver kit source code that can be open source friendly... anything. But it is possible that we did not get a reply or something similar of the last time with Sam Palmisano, but you know the expression " ‘No’ is Guaranteed ", so what can we loose asking again?
The kernel used should be the Linux kernel, not that its a great kernel but its actively developed, supports almost all hardware and there are tons of people who knows how it works. Implementing support for things like 16bit apps on a modern OS is not easy,
but using a really well known kernel with loads of developers would help.
I always loved the "Microkernel" idea and even the "Workplace OS" idea. I even think that the Android fragmentation problem was in part because of the Linux monolithic kernel (accompanied with the manufacturers that were so profit oriented and didn't want to compile new versions on the old devices). But I have to agree that the Linux kernel is widely used and possible the best option. I think it is the best shot because there is more knowledge of that kernel in general.
Regards