We've made improvements to our code related to the Graph API and have opened support cases with Microsoft.
The following is the response from Microsoft:
Microsoft Engineering investigated the logs and confirmed that the API call is failing due to a timeout from Graph waiting for the REST call to return data.
This is due to the number of items being searched, the size of the mailbox, number of recurrent items that need to be expanded, causing delays in REST being able to return data in a timely manner.
Our Engineering is suggesting reducing the size of the data being requested and reducing the number of calendar items being processed so the data does not timeout. In this instance, we regret you to inform this is a by-design behaviour.
While we strive to provide a comprehensive and satisfying user experience, we occasionally encounter limitations due to various factors, including technical constraints, resource allocation, and prioritization of development efforts.
Please be assured that we have carefully reviewed your suggestion, and although we are unable to implement it at this time, we have taken note of its significance and potential value for our users. Your feedback will be shared with our product development team for consideration in future updates and enhancements.
Microsoft Engineering would like to inform you that we don’t have a publicly published issue or bug ticket around this issue to track.
We apologize for the inconvenience caused.
We've observed that when using the Graph connector with a significant number of users (over 100) and a high volume of daily calendar sync operations (exceeding 30,000 item creates, updates, and deletes), some users may encounter the previously mentioned error, leading to delays in syncing.
If your organization has a similar or larger number of users and calendar sync activities, we recommend continuing to use Exchange EWS (Client Credentials) rather than the Graph connector for the time being.