You’re absolutely right.
As mentioned earlier (https://news.ycombinator.com/item?id=25993825), our goal is to be super transparent; we want you to fully understand what you’re spending on infrastructure. We feel good that there’s an incentive to help you work through the problems that you’ve mentioned.
Attributing collection and querying is made easier with authentication enabled by default. You can make your tenants as fine- or coarse-grained as you want, handing out authentication tokens to the producers writing to those tenants. This makes it easier to trace back to sources of bloat. You can also place rate limits on individual tenants to prevent bloat in the first place.
Additionally, we think users might reconsider the premise of the problem. Because the cost of running Opstrace follows cloud economics (because it runs in your own cloud account), it's basically as cheap as it can possibly be. So you might consider that you do not have as much pressure to curate what is stored as you think. (I didn’t say "no" pressure, but "less" might be a huge improvement. :-) )