Its actually a hardware bug that might be solvable with software. When the audio is "laser off", its at a different location than "laser on". Its a problem with the circuit, but we hope to get the software to get a calibrated offset whenever the laser is off. When the laser turns on at the same point on every layer, there is a small section that never gets cured properly because the laser wasn't there in time.
Until then I've used a bandaid fix that doesn't get rid of the twitch when the laser comes on, but randomizes the starting point of each layer, so it doesn't happen at the same point on every layer.
The program I made to randomize the gcode was my first useful python program and first time looking inside gcode files
It still needs hand editing to set the speed to start at (I probably should have left the first layer alone), and there might be other things.
Attachment 1356