It could finally detect if the transducer was running by reading the LED voltage, then press the button by grounding the button. This allowed the mist to be controlled with a switch. The mane problem is the LED has to be debounced for 3 seconds, in case it flashes. If the switch is toggled too fast, it takes a few seconds for the transducer to match the desired state. Fortunately, the switch's physical position is always eventually matched by the transducer.
When the 5 minute timeout shuts it down, it senses the LED going out & presses the button to restart it.
The water level sensor was well & truly corroded after just 1 day.
The next step was gold plated copper. Helas, this still built up a layer of scale. The problem is the water level constantly rises & falls, depositing a layer of stuff every time. The hope is the resistance eventually levels off.
The 2 wire sensor had enough variability to detect a low & high level, allowing for hysteresis. A bigger reservoir with more room for overflow & longer wires would get better results from a 2 wire sensor but cost more.
The automated power button & level sensor were good enough to finally design a 3 transducer unit.
An unfortunate discovery was raising the gatorade bottle by just a little or putting too much water in was enough to cause the water to siphon out & flood out of the sprayer. It might actually be more effective to have a solenoid valve instead of a pump & gravity feed it.
These things arrived. They all worked, contrary to reviews. This model doesn't time out & doesn't detect running dry, but they don't seem to burn out.
Through hole soldering is still essential for an experimental design like this.
It worked, with all the misters being managed automatically.
Current draw overheated the voltage regulator after a few hours.
The cable management is a mess.
The only reason for keeping all 3 boards was to detect being out of water, but now that some models don't detect being out of water, it might be worth making a custom board to drive all 3 with 2 wires.