I'd probably do the last one. Not sure how reliably it will be but will certainly be better than nothing

The biggest advantage to CAN bus is that it's a true multi-master bus. Any node can broadcast at any time. The other nodes will back off (there's a priority list for the nodes) and allow transmissions through. It's not a master/slave kind of system where nodes send/receive to addresses. A node broadcasts on the bus and the other nodes can decide whether to receive and act on the message, which opens up a much more distributed system of pub/sub style architecture, but with embedded systems.

The other biggie is that it uses a differential voltage transmission. Much like the difference between RS232 and RS485. A CAN bus may stretch 10's to 100's of meters, depending upon transmission speeds. With that kind of range, you can now wire up larger hardline sensor and controller systems for human-scale environments which I2C and SPI just can't do. It's also much more robust to handle noise on the lines, though it takes hardware level support for many of the features.

I haven't picked a ATTiny 10 programmer yet. For my ATTtiny 85 chips I have a Arduino shield that I wired up. For ATTiny 84 and ATMega 328p I have USB driven boards with ZIF sockets to make inserting/removing the microcontroller chips easier. Basically really cheap ones I found on Aliexpress years ago and they keep working just fine.

I use a combination of the ArduinoIDE and the ArduinoCLI to program the chips, depending upon the project and how much effort I want in my build toolchain.

Among the raw microcontrollers, the ATTiny 84/85 needs the least analog wiring support. The ATMega 328p needs you to provide at least a little external wiring.