Hi Charles,
thanks for the provided information.
Meanwhile, we have done a further code review of the relevant Graphics Engine source code, which handles the issue processing. From a logical software perspective the observed situation can not happen (as long as the GUI is running in one thread, what you already confirmed).
In the past we searched for a similar issue occurring after hours, which was caused by incorrect data read from cache. Have you considered the STM32H747 errata sheet - especially the section "Cortex-M7 data corruption when using data cache configured in write-through"? Please check your MPU configuration and let us know if this errata could explain the observed issue.
Best regards,
Manfred.