There seems to be a requirement for high quality DNA. But if run with herbarium specimens maybe sequence capture and assembly are poorer relatively but adequate for phylogenetics with a reduced number of loci?
Elsewhere I got information about an article related to: Hamon et al 2017
Genotyping-by-sequencing phylogeny for coffee (Coffea)... Here there were 13 herbarium specimens used. The herbarium specimens were considered good if they were relatively green, and had untreated adult leaves. During DNA extraction gels were run and a clear unique band of DNA at 20 kilobases indicated likely success. Apparently ~40% success rate was obtained for another study by those authors.
Dear Chris, if you are doing GBS the number of loci you will get is anyway rather low for phylogenetic studies on higher levels (compared to (dd)RAD seq) and, because the restriction enzyme recognition sites might be lost due to mutations in not closely related taxa. However, using herbarium material will even increase this loci dropout due to bad DNA quality. Our experience is that it might work for well-prepared herbarium specimens that were dried fast, but you will receive anyway much less loci than with fresh material. Of course you can deal with missing data, but if the amount of gaps is too big that causes biases in your analyses. The same is true for too less loci (e.g. on a broader phylogenetic scale in orchids we received with fresh (!) dried Silica material just 18 shared loci by using GBS, but it worked fine on a species complex of a few closely related taxa). Depending on your research question, especially if you have just herbarium specimens, I think GBS might not be the best method.