From reading the docs this service seems

From reading the docs this service seems most useful for looking at trends and data points at a big picture rather than being optimized for accuracy. That makes sense for lots of data points and general analytics, but what's the suggestion for things that require higher levels of accuracy that may not tolerate sampling? For example, what's the suggested way to store view counts? Thinking a durable object may be the best bet but that doesn't give an easy way to do time series data.

Alternatively could use D1 to just store it with SQL, but if there's a best practice or way to use the platform that integrates with analytics engine I'd prefer to go that route.

Thanks for your thoughts!