Multiple devices fail to upgrade

We have seen a significant amount of devices (~10%) that fail to upgrade. While it is possible that it is a network problem for some of them, even some of the devices in our test setup fail to upgrade. Some of these versions are running on a version with a known bug, so customers are complaining.

How do we go about resolving this?

Are these devices to which you have SSH access, then? Do you see the failed upgrade in snap changes? You can get further information on the upgrade with snap change <change id>, which may prove helpful to determining why the upgrade failed.