I have been doing exactly this recently. Though my CO2 sensor was 3.3V. But a solution could be powering the sensor with the VIN 5v pin and using a level shifter for communication. That's what I did with the 5V screen I used (though I didn't need a level shifter). I also added a lithium battery with a charging and safety circuit hooked up to a boost converter making it 5V and connected that with VIN and GND.
In my use of bistable latching relays, where the state of the relay position is information that must be tracked, I have found it to be easiest to use an optocoupler (H11AA1 or PC817C) to feedback the state of the relay position, which is polled at bootstrap. If you have multiple states to read at boot, you can use something like the TCA9548A multiplexer to accomplish the state read and to set your volatile state variables appropriately, and save on your μC pin count.
@fifo Well done! - probably the new module needed a little extra power and the power brick was already failing so this pushed it over the edge. Failing psu can be difficult to diagnose until they fail completely, especially if encapsulated.
Power supply and cabling are usual suspects in this type of work.
Well done and good luck with the rest of the project!
By the PCB silk I was wrongly thinking that the module could be on of "RFM69" or "RFM69H" or "RFM69W" and "RFM69HW".
Now if I understood correctly the diagram, the module can only be either the "RFM69W" (low power 13dBm) or the "RFM69HW" (high power 20dBm).
Since this module has the "H" marked it should be "RFM69HW" and that is confirmed by the other side of it, because it has the two additional black ICs.