ESP32 haptic firmware vs laptop cue scheduler: check single-motor state, pulse cap, and cap vs longest cue before the first hardware run
Rule: when a laptop schedules multi-step haptic cues and an ESP32 executes PULSE <motor> <ms> lines, diff the firmware against the laptop cue table before hardware arrives. Three bugs found by reading only: (1) firmware kept one activeMotor and called stopMotor() on every PULSE, so cues with simultaneous steps (both sides at offset 0) or overlapping motors are cut to the last motor; use per-motor deadlines with a small concurrency cap for a 2xAA rail. (2) firmware maxPulseMs=300 silently dropped longer PULSE lines while a laptop cue (lost target) used 600 ms, so it never buzzed; make the caps agree (laptop link cap 1000 ms) or split the cue, and log rejected lines. (3) the integrated firmware sketch had only PULSE/STOP, no MIC stream or BTN events although the laptop already parsed them; and mic pin/gain tables differed from the protocol doc (right/left swapped) so keep serial field order fixed and move each gain with its channel. Method: python one-liner printing max ms and motor sets for every cue in haptics.py, plus grep of the sketch for Serial.print/MIC/BTN.