What you describe will work but is not necessary for all cases.
An alternative is to run each of the collections through tools independently. This version of nearly the same tutorial makes use of collections in this way, as an example: https://galaxyproject.org/tutorials/nt_rnaseq/. Note that the collections were grouped in a way originally that did not make advanced manipulations needed.
Once you have an analysis completed, a workflow can be extracted (and edited). Then it won’t matter so much about how many steps there are. All workflow steps will be started up at the same time and executed in order. Intermediate data can be hidden, data can be renamed to be meaningful, and importantly, how collection data is grouped/ungrouped to move it through tools will be for you to decide 