I'm using the pipeline to analyze bacterial RNA-Seq data. Even though ribo-depletion was included in the library prep for these samples, there is still up to ~10% rRNA content reported by SortMeRNA. I wonder if people always turn on rRNA removal for bacterial RNA-Seq (or even any species RNA-Seq)?

I'm not sure there is much benefit in removing rRNA prior to mapping, except that it maybe makes some of the QC look better. You are just double-handling reads -- first running all reads through mapping against rRNA, then again the remaining 90% against the genome, I don't see any benefit, unless the first step is much faster and you have a lot (>50%) of rRNA. The rRNA does no harm, you can always remove these genes from the counts table later in case they mess up your normalization, or something (they should not, in my experience).