well this has been an identical ongoing issue for me for quite some time.
first and foremost: Disable moving faults. its the only thing that can quell the terror. you just have to dump it on its side when you dismount. elegant? no. do you have to like plan your dismount always? yes, but the piece of mind, priceless
my gear: lil focr 3.1.1/18s2p 50S/exile sensor
this began sometime after installing exile. (and actually, compared to the ghosting that occured after install until this began, this is hella preferable [one time: board ghosted in reverse after i stopped at a stoplight in the bikelane. board was 3 blocks away by the time i noticed it was gone, continued on, up and then down a hill out of sight a 1/4 -a 1/2 mile before its course caused it to just veer ever so slightly into a curb. another time: dismounted on a grass island between four lanes of traffic, i look back as it speeds into oncoming traffic in rush hour, board is hit by the front left tire of a chevy excursion, flipping it several times end over end, shattering FM fender and snapping my badger torque bumper in half. ran out snatched it, then ranback to the other side of traffic and jumped on it and sped away as the SUV came to a stop, but was unable to get to where I was.])
anyway, as far as i can tell, the sensors are 4 layers: top plastic, bottom plastic, and two layers of metal contacts layered in between. im pretty sure that these layers, the top and bottom plastic layers, become unsealed to one another, and thats the wiggle that causes the sensor to get twitchy during engagement. prolly doing flywheel with your front footpad downput shear force on the edge of the sensor and caused it to delaminate.
has nothing to do with imu, imu calibration doesnt read or modify adc values as any part of what is essentially a gyroscope calibration. and i bet you didnt ride the board between imu calibration and flywheel, so causation in that respect cannot be confirmed. also i noticed adc voltages used to be pressure sensitive. and the voltage was proportionate to pressure, but somewhere along the line its on/off. the slightest pressure will activate the full adc engagement value. mine was 3.27 and 2.06, respectively. so as long as your values exceed those amounts, even by .01 V, you can never fail to activate the required activation threshold, and lowering them to, say 1.0 V is indistinguishable from any other value less than what your ADC’s post at.
things that have helped a little: heat up grip tape all over, and wack it with the heel of a shoe (exile tip for griptape expanding and shrinking over time due to heating and cooling and use);
things that has helped drastically. heat up and remove grip tape completely, and then heat up and remove very carefully the entire sensor and reseat it. the exile sensor like adveritsed like “branch it out,” to cover the most surface area and reach the far edges of the platypus pad… but the sensor is cut flat and if your stretch it out, youre putting a slight curve where it doesnt lie flat somewhere. the sensor cant work right like this and the slight ripple creates surtace tension that either could create the initial layer shearing and or exacerbate the tweakiness of an already sheared sensor. two complete sensor reseatings have fully fixed one of my adcs to expected, and the other one is id say 80% where it should be…. almost to the point where i could enable moving faults, but i dont want half state faults at below half state erpm … not cut its dangerous per se, but moreso because you look like a dumbass when it happens
oh also, half switch and fullswitch fault delay: bump up from 250 ms to 500 or 1000 ms, so your twitch has to cut out longer to cut and set the half state fault erpm down from 200 to 100 or 50, so if you are in motion at all, it wont cut, and you have to be at a dead stop to activate heel lift
AND fuck this recalibrate nonsense. reflash firmware often and take it from 0 to fully calibrated from a blank slate. i cant explain it, but this has frequently solved issues that were inexplicable. youll learn the settings quickly with repitition, and never restore settings backups, ir youll be flashing back the same motor shudder/whatever you flashed to try and fix. there was a time when my board would get good numbers only on a fresh flash, and sometimes id have to reflash after a power cycle cuz it would idk just become sludgy and buzz otherwise. but every fresh flash itd work like a dream. there was no setting change that could explain it, and id use the same exact settings, but those settings would only feel right after a reflash. and this last part might be in my head cuz i know the vesc verifies crc’s after bluetooth firmware transfer and before it executes the actual flash, but, ive hD statistically significant results interms of overall stability flahshing firmware over usb rather than bluetooth. YMMV
if i am correct this can be tested by swalpping front footpad with any correctly engaging vesc or future motion board, if its fixed, then get a new sensor or find a way to glue the top and bottom plastic layers on just the very edge/corners. if you open it up too much the sensor will be shot à la cutting/tearing a stock FM sensor during stock footpad to kush high transplant operation. if you break or cut the actual circuit in the intermediary layers, its shot. repair is probably
alternately….. my last stock xr build before i went all in on vesc, after fucking up to pint to xr sensor transplants in a day (the only sensors in town, and 100% of available stock (and $300 total for two stock future motion full footpad sets for pint) insaid fuck it, and i drilled a fucking hole through the xr front footpad, and wired the adc switchcraft pigtail to a 2 way MO physical button switch that poked through. and though it was very quickly vesced over, i never had a single undesirable ADC fault or engagement issue with this setup.