Do you use the same codebase for all editions? Meaning, mongodb on all editions? Doesn't mongodb explode in disk-usage (no compression, field-names etc)?

Since everything is dynamic/realtime/no-batch-processes, how do you handle aggregations? With just group-by all the time, or keeping most requested dimensions with counters and other with on-demand-group-by?

Does on-demand-groupby get slow with big-customers?

If you do keep counters (example: daily revenue), what do you do to keep things consistent(so in then end you have the same number by reading the counter or doing group-by on raw data) (ie: no mongodb transactions) ?

Does mongodb aggragation-capabilities satisfy your app or do you need sometimes to group-by-in-code?

How do you count the "online-right-now" users?