The Reference Sequence (RefSeq) collection provides a comprehensive, integrated, non-redundant, well-annotated set of sequences, including genomic DNA, transcripts, and proteins.
I think the above is quite buzzword-heavy and does not explain what RefSeq actually is. What does "comprehensive" and "integrated" really mean? I am not sure.
In a nutshell, I believe that Refseq curators take entries submitted to Genbank and designate one GenBank entry as a representative sequence for each organism (this is the non-redundant part). They probably pick a representative sequence that is well annotated. But frankly, I could be wrong. I am assuming this is what happens.
For example for the millions of SARS-COV-2 sequences out there, there is only one "reference" sequence, that also has both a GenBank and a RefSeq id, for the exact same sequence.