At a certain level of aggregation, metadata is AWS need-to-know for capacity planning etc or for account execs to understand the needs of their covered clients.

There's definitely room for discussion over when what level of granularity is necessary and whether metadata at different levels of granularity should be shared with product vs. customer facing teams.

That didn't seem to be what the GP comment was about though - 'mining data from AWS servers' sounded to me like a much more invasive approach to client data.