I experienced this too on the weekend. Cutting great until it didn’t.
The Top Left belt in my case simply didn’t retract during a part near the end of a cut, the software stopped the job after a 15mm deviation was detected, saving my material and preventing damage to the machine. Then it wouldn’t retract fully during “retract all”.
Upping the retraction force to 2700 almost worked, but I still had about 10cm hanging out.
I took the machine and encoder apart and found nothing wrong except the belt roll was tightly jammed against the idler and the grub screw on the motor pinion was loose.
I slid the loose pinion off the motor, manually rewound the belt tightly and saw it was now about .5mm less tight against the idler.
I loosened the 4 bolts holding the motor and pushed that as far back away from the idler as I could and tightened the bolts back up.
I tightened the grub screw like heck, although i doubt this would make much difference.
Feeling silly, and wondering if I should solder the eoncoder board pin or 3d print a belt guard, I decided against it and put everything back together.
The machine worked perfectly once re-assembled, and I was able to “retract all” on all 4 belts with 1500 force. I was then able to resume my cut where it left off and still got a usable part after all that. 
Sounds like the improvements to the software over time have spared me a lot of issues.
Edit: maybe tightening the grub screw is important - to avoid the possibility of a jam on an unseated bearing, per this thread: All belts retracting but one