Being a simple computer monk, my choices are based on simple factors. The following paragraphs help explain these.

The choices include the long term ongoing usability of each system. As such true Open Source is helpful for each, so is the ongoing use of Assembly Language on each. Then there is the ability to support dates into the distant future (hence the lack of Macintosh System Software now). Then there is my personal familiarity with each system, a definite must.

AMIGA:

The Amiga is still widely used, uses a date format that allows for the next 11.7 million years before roll over. Also this is a simple widely cloned OS, that is easy to program for and well designed.

Both Assembly Language and BASIC remain fun to use on the Amiga OS. With most BASIC implementations being extended and compiled AmigaBASIC compatible systems, and very capable in nature.

RISC OS:

RISC OS has a few thousand years on its date format. Also RISC OS is easy to program for, fun to use, simple enough to completely understand. Then there is that it is the only OS that is well tied to the ARM 32-bit ISA.

There is also the fun of using BBC BASIC V, including its built in assembler. As high level languages go it is a very fun to use one.

Also is the fact that writing simple (in my case very simple) emulators for RISC OS is much more simple than any other system I have ever used. This is for emulating Systems that use a different CPU from the host, virtualizing the 680x0 is very easy (except if the host is a 68000, minimum of 68010 for virtualizing), as is virtualizing the ARM (up to ARMv6, ARMv7 and 8 have some bariers to virtualization).

Atari 68K:

To be honest this one is just for its simplicity, ease of programming and personal attachment. The date format is too limited for real long term use.

There is also the fun in playing in C that is well suited to Atari TOS and clone OS's. Probably the only non-nix OS that I have used where C is a good thing.

There is also the choice in Multitasking enviroments, with a few very good Cooperative Multitasking options that outperform the Preemptive Multitasking options.