CDN edge caching for APIs: how aggressive can you be?

we're exploring more aggressive cdn edge caching for our product catalog api, which sees around 100k requests per minute. the data changes about 100 times a day, so it's not super volatile, but not static either. the challenge is figuring out how aggressive we can actually be without serving stale or incorrect data. we're dealing with issues like authenticated requests that can't be cached easily, designing cache keys that are specific enough without being too granular, and figuring out effective purging strategies when data does change. personalized responses are another headache. has anyone successfully cached a significant portion of their api traffic at the cdn edge for non-static content? what were your biggest hurdles and key lessons learned, especially around cache key design and invalidation?