It is expected. Ties in the FDR are a property of the Benjamini-Hochberg method which edgeR (and limma, DESeq2, pretty much any OMICS tool) uses for multiple testing correction.

The fact that FDR is not strictly monotonic in the p-values, leading to runs of equal values, is the reason why I have always advised people to make volcano plots using -log(p-value) or log(posterior odds) for the vertical axis instead of -log(FDR), see for example:

• https://support.bioconductor.org/p/9141879/
• https://support.bioconductor.org/p/98442/
• https://support.bioconductor.org/p/62384/
• The FDR and P value in RNA-seq analysis

See also the help page for volcanoplot() in the limma package.