Everyone understands how useful it is to capture a user's workload via mongoreplay and play it back in a controlled environment with mongodb to test new builds. The problem is that users may be reluctant to share mongreplay recording due to security and privacy concerns.
If we had an option to obfuscate the user namespaces, index names, view names, field names and values in the captured files, it would make it easier for users to share capture files. We then could test our latest builds against their specific workloads.
More forward looking is to then build up a library of workloads from various users. We could also have a tool that analyzes these files and generates statistics: 3000 collections referenced, 2000 collections inserted into, 1000 collections with finds, 30 with aggregations, 800QPS, 6K inserts/sec, etc. We could pick sets of workloads from the libraries to test new builds.