I think I can help shed some light on this… NRDB is schema-less, which means that any event/metric type can have any named attribute. Querying follows this pattern, so it’s possible to look for any attributes on any event/metric. NRDB will treat this as a genuine query and will attempt to look for any attribute/s matching your query. The autocomplete function in Query Builder is designed to help you avoid typos, or you can use Data Explorer to visually find attributes and build your NRQL queries.
If you are familiar with relational databases this seems like a strange concept, but it’s for good reason - the flexibility of NRDB means you can add custom attributes on the fly and don’t need to worry about data migration, normalization, etc. You may be interested in this white paper which goes into more detail: https://newrelic.com/resources/white-papers/inside-nrdb-flexible-unified-database