Dumb question - 

If a rollup switches out DA to some low cost DA solution, is there a major difference in cost savings if it’s ZK vs Optimistic? What’s the best way to think about the differences there?

Optimistic rollups need more storage, because they need to publish enough information for anyone to replay the transactions. Validity rollups only need to publish the state differences, and a short proof that transition followed the rules. It's a tradeoff of compute and storage. Thus saving is larger for optimistic.

Do you have a sense for how big the difference is?

No, but it's going to be a range. How good is your compression, what types of transactions, how often is an update published etc'. In theory, there is almost no upper limit to the savings on a validity rollup. A million trades can collapse into a handful of balance changes in the best case.

There was a post (by Vitalik if memory severs me right) about the savings and data per tx type in optimistic rollups. It would be a good starting point for any calculation.