Currently, the Canvas SIS integration rostering feature will import data collected from your SIS as-is without further validation. This is generally desired and expected for a scheduled unattended synchronization. However, some unexpected conditions at the source SIS could lead to the collection of errant or incomplete data, and this could lead to mass unexpected change in rostered Canvas data, or even deletion of data such as enrollments in courses.
The SIS Integration could implement a freeze or hold feature that waits for approval by an administrator when the collected data represents a significant percentage of change. Since these imports are already done in diffing mode, I believe this would feasible to achieve. Many applications already support a similar feature for rostering from SIS data. Thanks for considering this suggestion!
