The change in ZBXNEXT-9686 tries to gather the return code of sudo. But sudo also passes the return code of smartctl through.
smartctl has many different return codes depending on the state of the disk. So it's not always an execution error.

With current Zabbix Agent2 versions smart.disk.discovery works only if all disks are healthy. If there is one disk with smart errors than the whole discovery crashes.

 
Working with 7.0.8: smart.disk.discovery_working_7.0.8.txt
Broken since 7.0.9: smart.disk.discovery_broken_7.0.19.txt

 
2 examples with smartctl outputs of the same host as above:
disk healthy: smartctl_disk_healthy.txt
disk with issues: smartctl_disk_with_issues.txt