Hello and thank you for posting on our forum!
Here's a historical case that might help you:
[Analysis]
1. Checking swichover methods, the customer used a power-off server to switch;
2. When MCU dropped, all sites on the MCU dropped;
3. Checked the dropped MCU and the normal one, found the media resources were different.
[Root cause]
Installed MCU has a different MCU resource that couldn't fulfil the swithover test.
[Solution]
After reinstalling MCU and making the resource balanced, the problem was solved.