I tried the suggested change and tested. But it did not provide the expected results.
If a on was requested while an off was still pending to be processed, the off-request flag gets cleared without the actual request being processed and the off request was never executed.

Not sure these min/max are right here either.

So we have all temp sensors getting averaged now, right? So should we have one get function that takes a sensor ID as param?

Remove extra blank lines.

Remove this reference (across the file) if the implementation is updated.

Update the deadband timing in the comments.

Remove blank line. Also, add comments on the key logics for reference.

Default case rise count should be 2, instead of 1?

Now that we go with pressure slope detection logic, we don't want to use pressure-based detection for Low Qd. correct?

The BC valves not even opened and collecting the spent dial pressure as last previous spent dial pressure may not be correct.

Done.

Done

Done.

Same comment.

Having much larger P coeff is likely to cause a lot of oscillations and very slow to get to stable target.

Input is d78TempAvgC

I don't think these min/max (for control output in output units i.e. PWM) are appropriate for limiting maximum error sum step size (error is in feedback/target units i.e. RPM).

Done.

Just following the current style of code implementation to keep it uniform. These defines are temporary and will be removed.

Done.

Done.

Done.

Done.