[dub]

The reasons Vitess didn't have foreign key support historically actually weren't a bold performance tradeoff or anything like that. It was more of a classic, boring, backlog prioritization thing: everyone heavily using Vitess was using gh-ost for schema changes, and gh-ost didn't support foreign keys.

Now that Vitess has native schema change tools it's more reasonable to revisit user-friendly, out-of-the-box foreign key support.

[samlambert]

Kind of. It was definitely an opinionated choice at some point, justified with performance. Now it is just something we need to add and it's nearly done. It will still have trade offs associated like all distributed systems.