Hi @MichelaRiba! Okay, time for that update that I promised. I started working on a simple analysis using the dataset provided here (link). First, I did a de novo clustering - it was quickly done. Then I moved on to closed-reference clustering - I observed the same issue you found - vsearch is complete rather quickly (maybe 10-20 minutes), but then something is hung up and going crazy slow. I began a detailed debugging, and identified a huge performance bug in the q2-vsearch code. I owe you an apology - thank you for sticking with it! Basically, the bug only presents itself in certain cases - namely when you have many many many unique features, before clustering. I have submitted a patch (link), and the current plan is for this to be included in the upcoming QIIME 2 2020.5 release (link). I think at this point there is no need for your attempt to reproduce. Thank you again for your perseverance, this was a hard one to diagnose and reproduce! Thanks!
:qiime2: