Consider counting attributes on data and tracking as `meta.attribute_count`

[cartermp]

Feedback from CS: this would help them a lot since they can measure how many fields teams usually use, but they don't know the total. Could also be other ways to get that, but it seems possible to count up on ingest.

[cartermp]

something something sampling could complicate?

[kentquirk]

Sampling doesn't affect the set of attributes within a span, only the number of actual spans.

A meta. attribute is something that would decorate an individual span (i.e., in the customer's data). We could do it, but it requires sudo access to customer data to see it. Is that the desire here?

I don't actually understand the original statement "they can measure fields but don't know the total"? Total of what?

Please be more specific as to what's wanted.