Versioning is most useful when you know what recovery looks like. Start with a disposable object in a bucket that already has versioning enabled. Write two revisions, record the checksum of the first revision, and recover it to a scratch path.
aws s3api list-object-versions \
--bucket example-bucket --prefix reports/demo.csv \
--query 'Versions[].[Key,VersionId,LastModified,IsLatest]'
aws s3api get-object \
--bucket example-bucket --key reports/demo.csv \
--version-id "$VERSION_ID" /tmp/restored-demo.csv
sha256sum /tmp/restored-demo.csv
Make the verification meaningful
Check that the version belongs to the exact key. A prefix can match several nearby names. Compare the restored checksum with the recorded value, then open or parse the object. For an encrypted object, the recovery identity may need access to the encryption key as well as permission to read the specific S3 version.
A simple delete in a versioned bucket normally creates a delete marker. An older object version can still be found by version ID until a retention or lifecycle policy removes it. Check how long noncurrent versions are retained and practice with an object old enough to represent your recovery needs.
Record the time it took to select, recover, and verify the version. That time is far more useful during an incident than a note that says versioning is enabled.