These days read trimming is generally unnecessary in both DNA and RNAseq. Is there any reason why the same shouldn't be true of bisulfite-treated WGBS data?
Yes, we typically trim reads for WGBS even when using local alignment (bwa-meth in our case). The primary reason is that coverage for WGBS tends to be low-ish and we don't want to chance parts of the adapter getting aligned and used for methylation extraction. With fastp and recent cutadapt changes (it's finally multithreaded) at least the trimming step doesn't take seemingly forever anymore.
