Hi @Biancabrown, sorry it took as longer than usual to reply!
We have been actively researching the dada2 slowness when installed in a conda environment, but unfortunately we don’t have an ETA on when these changes will get rolled into QIIME 2 
. In the meantime, it does seem like the approach you proposed here could work just fine!
Looking at your code snippet, now that you have your feature table, you can use that to produce your representative sequences. We have some python code in q2-dada2 that does this exact same thing - it reads in the BIOM table, does some ID cleanup in the table, then extracts the representative sequences. I am not an R-aficionado, so I don’t think I can be of much help here when it comes to writing some comparable R code, but hopefully this can get you moving in the right direction. Please keep us posted if you get stuck! Thanks! 