My dataset contains mini-bulk samples (5 cells per sample). I have 6 replicates for each cell subset, meaning that I sorted 6 wells with 5 cells/well per subset. I've analyzed my data using the 'regular' DESeq2 pipeline and was wondering if I should apply the recommendations for single-cell analysis from the vignette, because of many zero's? Any pro- or con- arguments for this?

I can't tell with this information whether it would be useful to apply ZINB-WaVE here or if the standard approach would be better.

scRNA-seq count distributions vary widely by protocol.

I would recommend that you start with the standard approach, and examine per-gene results with plotCounts. You should look at genes that you expect to come up (e.g. known markers), as well as genes that are significant but maybe have low number of non-zero counts for example. If you see either suspected false negative or false positives, consider alternative approaches, including ZINB-WaVE or other approaches compared here:

https://www.nature.com/articles/nmeth.4612