I would like the ability to set tags on individual nodes so that I can change the read preference with more granularity.

Mongo supports these tags on non-Atlas deployments: https://www.mongodb.com/docs/manual/core/read-preference-tags/#std-label-replica-set-read-preference-tag-sets

Analytic Nodes are very useful, and they are built as way of isolating workloads. The problem is that you can only really isolate into two different buckets (well three if you consider local read nodes). What I'd like to do is isolate more than two different sets of users and queries, so say user reporting and internal dev activities have their own sets, and they work independently.

You could kind of achieve what I want, by just letting me configure multiple sets of analytic nodes and then give each one a custom name, that appears as a tag, that I can use the readPreferenceTags option to select: https://www.mongodb.com/docs/v3.0/reference/connection-string/#uri.readPreferenceTags