When the HTML parser in CommonsMetadata is set into multilang mode, it will return an array of language versions for each metadata field, one for each language template ({{en}} etc) used in that field. But for fields not using such templates at all, it will return a single value, not an array. This makes sense for fields which are not language-dependent, but for the rest it just makes the behavior of the extension hard to predict.
Version: unspecified
Severity: enhancement