Sometimes the dbSNP team originally thought that it really were two different SNPs and thus assigned two different rs numbers which were merged in later builds.
Also see the answer to this question: [?]"How do I query dbSNP so that it will return a flat or xml file containing the new RefSNP (rs) ID number into which a previously valid rs recently merged?"[?] in the SNP FAQ archive book on the NCBI bookshelf. It gives links and information on where you can find [?]the archive[?] of merged rs numbers and how to deal with those.
The answer from Chris is correct. One reason for duplicate rs numbers arises from the fact that SNP discovery work started before the genome was "complete." SNPs were mapped to the Golden Path - and that simple statement does not begin to divulge how hard that work was. Nonetheless, as genome contigs merged into segments representing chromosome arms, so did SNP rs accessions merge into a single entry.
Back in the days of single-gene or single-cDNA sequencing, it would not be uncommon to see that two different groups each submit to GenBank their sequence data for the same gene. Each would be assigned an accession number but each represented the same gene.
Sometimes I've seen SNP that could be described ambiguously split into two entries.
For instance, one SNP will be described as AGTCC[C-], and a second SNP is described as AGT[C-]CC, and positioned two bases away from the first one.
Yes, proper form dictates that the former is preferred, but if the latter was described properly in reference to a transcript that runs the other way, it's an easy mistake to understand.
This can be quite common in dbSNP and in version 129 there was a huge surge of such entries (probably due to automated assignment of rs No.?)
Our tool (http://pfs.nus.edu.sg) can help to clarify this problem. Just search the rs No. and all the other rs No. pointing to the same position would be listed as "Alias" in the search result. (e.g: search for rs2032582 will give you 2 entries.)
You can also use http://pfs.nus.edu.sg/handytools.aspx if your know the gene context of your SNP. Just specify the gene context and all rs No. referring to this context would be given.