I replaced my aging IKEA Trådfri hub with a SONOFF Zigbee 3.0 Dongle-E (V2) and re-added all my (mostly IKEA) lights and other devices (41) altogether. This same error is now a constant occurence. For example, trying to adjust 4 blinds, or turn a group of light bulbs on or off, it’s likely that one of them will not respond, and the UI will pop up the “failed to deliver message” error. This doesn’t appear entirely like a sporadic single-message delivery problem - retrying the same action, it’s likely that the same device will fail again, until some later time when the mesh may have reconfigured.
So not only would implementing a retry logic in every automation be a crazy effort as @Markus99 wrote above, it probably wouldn’t even work. There’s something wrong at a deeper layer of mesh network maintenance.
The HA+Trådfri combo had its issues, like the tradfri integration losing connection and requiring a reload very frequently, but this behavior of losing zigbee connectivity to single devices is new to me.