It can happen that AcoustID does not have full MB metadata for a matched recording and only gives back the recording ID. It is assumed this happens if the mirrored MB database used by AcoustID is outdated. The response from acoustID can look like this:

{
    "results": [
        {
            "id": "674f06c2-4135-499b-9daa-03724acb5ac7",
            "recordings": [
                {
                    "id": "837a0c07-9da1-40fa-953c-8f1f58c4187f",
                    "sources": 52
                }
            ],
            "score": 1.0
        }
    ],
    "status": "ok"
}

The result only gives ID and no. of sources. The otherwise present fields "title", "artists", "duration" and "releasegroups" are missing.

In such a case Picard currently matches the file only to the recording ID, resulting in loading a standalone recording.

Instead Picard should perform an additional lookup for the recording ID against MB.org to retrieve the missing metadata.

See also discussion on https://community.metabrainz.org/t/picard-scan-still-returning-new-acoustid-submissions-as-standalone-recordings/603676