I’m here after 20 years of Homeseer with IO mostly via Arduino+Ethernet shield connected to relay boards at the moment.
I’ve always been a little uncomfortable with all the logic and control being done in the PC. I’ve had it once before where the software has gone down after chlorine has started pumping into the pool, but before the stop command has been sent.
So I’d like an element of local control, and I started using HA a couple of weeks ago.
My primary HA instance is a docker on Unraid, with NodeRed, and MQTT dockers to support. I think most of my complex mashup automation will be done in NodeRed.
I have replaced my heating controls with a NodeMCU running ESPHome which has given me local control backup and watchdogs if HA is not up.
My pool area logic is more complex with several temperature inputs, zone valves, heat pump control, solar thermal, chemical pumping and lighting control needed. I’d rather build that in NodeRed than ESPHome. I also want to use wired ethernet rather than wifi.
So would it be best to put a “slave” HA PI there, in supervisor mode with NodeRed and subscribing to my main MQTT instance? I can use the GPIO on the PI to drive my relays, and the 1wire temp sensors.
I will have about 10 virtual switches on primary HA that need to be subscribed to on the Pool Pi and used in the logic. I’ll want to push the temperature and valve statuses from the PI back into the main HA. It is straightforward to build linked switches and sensors in the ESPhome integration, but what is the best solution with 2 HA instances? Do I even need HA installed on the slave, or should I just use NodeRed?
Any suggestions much appreciated 
Cheers.