As I was writing the details up I couldn't help but think of my other flat and tiny node, which really is more of a human interface device.  It is almost the same as this switch setup without the switch, and with an OLED screen instead.  

I described it more in detail here.  It uses a LiPo instead of a button rechargeable, and has a JST connector for easy recharging The screen node doesn't last very long, hours compared to potentially years with the low-power programming of the switch node. Other than this, there are minor mechanical differences, like it lacks a programming header, though I did not have to add that to the switch setup, I just did that for convenience.

This unit is location-aware and can re-purpose its displayed information dependent on its location.  I was thinking, I could use the exact same code in the switch units, which are not coded to be location aware, they are just functionally assigned.  I was further thinking that maybe the thing to do here is to merge the code, and specify whether or not there is a screen, or a switch, and whether or not it should be location-reassignable, or functionally assigned, or some combo.

Maybe that is too complex.  The nodes themselves are so simple that it is almost worth it, to just maintain one codebase instead of two.  I'm not sure that I would ever want or need a location-reassigning switch though.  

Now that I have said that, I will need one soon, I am sure.