Supervisor needs to start the containers it manages for addons, plugins and Home Assistant in the correct order after a system reboot. Changing the restart policy it sets on those containers may cause them to start in the wrong order and create errors.
If the restart policy of observer was changed, fix it from the host shell with this:
docker update hassio_observer --restart always
For everything else, the restart policy can be fixed with the following command:
docker update <container_name> --restart no
The supervisor log should contain a list of container names with incorrect restart policies.