One more thing on A.C90 that bit us last year — check your encoder cable routing relative to the VFD power cables. Yaskawa spec says 200mm minimum separation, but on a press retrofit the cable tray is usually shared and nobody measures.
We had an intermittent A.C90 that only showed up above 30 SPM. Turned out the encoder cable was zip-tied to the motor power cable for about 400mm inside the cable chain. At low speed the noise coupling wasn't enough to corrupt the serial comms, but at higher speeds the dV/dt from the PWM switching was enough to flip bits. Rerouted the encoder cable to the opposite side of the cable chain and it's been clean for 8 months now.
The diagnostic trick I use: Yaskawa has a monitor parameter Un00E that shows encoder communication error count. It increments even when the count isn't high enough to trigger the alarm. If you see that number climbing during production — even slowly — you've got a marginal connection or noise issue that will eventually become an A.C90. Check it at the start and end of each shift for a week. If it's climbing, start hunting before it becomes a production stopper.
Also worth noting — if you replaced the encoder cable and the alarm persists, check the encoder 5V supply voltage at the drive end (Un012 on Sigma-7). Should be 4.75-5.25V. We had one where a corroded ground pin was dropping the 5V supply to 4.4V under load. Encoder worked fine at rest but would glitch during high-speed moves when current draw spiked.