You sound like me in that you run multiple systems. I typically run A3000 w/ OS3.9. I will be reconfiguring the 3000 and going back to a basic modified Workbench 3.1. I'm not sure if I will even run Scalos on it. My best system stability came from an updated 3.1 configuration. I only run a few specific applications on this machine so having every latest OS gizmo and capability not an interest to me.
Regarding my Morph/Efika setup. It has been purely experimental. I do not use the machine for anything useful yet. I have 2.2 and am debating on 2.3. Overall the system runs good despite the severe hardware limitations. I have been snooping through a lot of Freescale documentation concerning the 5200B and am seriously considering modifying my board by dumping the 44pin header and replacing with standard 40pin in order to get my CDRW onto something faster than USB1.1. Despite the crying I have heard about this modification and that ribbon cables are impossible to use, Freescale documentation clearly indicates no problem below 18" length. 5200B Lite dev board is set up this way. I'll let you know how it turns out. The real interesting thing about Efika is its memory system. It is actually quite fast. I think Gunnar once eluded to this within this forum at one point. I read up on 5200B and chips used within Efika and agree what Gunnar said about Efika's Ram.
I have not engaged in OS4. Not sure I will yet. I do like the SAM board for the sheer fact that it is simple, has some expandability, and has reasonably modern features. OS4 seems to have a fairly active developer scene so this is why I still remain interested.
My Aros experience is limited to running it through VME. I may attempt a dedicated install on older PC we just pulled out of service at our house. I believe the latest Icaros now includes a working Poseidon USB stack. Aros has always been interesting to me. We all obviously use AFA in our 68K machines. I do not know of full 68K porting requirements for Aros. It seems to me that it comes down to original chipset support. Many people still use chipset dependent programs. I would tend to think that Aros porting efforts in this regard would be difficult. For people like myself, the programs I typically use support RTG. So a generic port of Aros to 68K would be beneficial to me.
Obviously with Natami, it makes more sense to use the 'improved' chipset because it will run very well. This makes a full AROS port (and one taking advantage of Natami) seem far fetched to me.
Wow, I'm really rambling here so I'll stop...