I would like to see the granularity of the configuration reduced from a global or an application level to the object level, together with the ability to archive data retention and make reporting and indexes faster. I'd also like the ability to get inside the individual information collections on a more specific level and to get information directly related to a topic. They currently have BIRT, a business information reporting tool and I'm forced to use that to extrapolate information unless I want to go through the command line which has very poorly documented access. They put out a syntax and then you have to experiment to figure out how it actually works. One area of improvement would be the documentation on the application, its third party dependencies, and the functionality of the command line features. I think improvements could be made in performance. I have some situations where I can throw more resources than the application can handle, and I will not get better performance. The issue is that indexing is single-threaded so I'll never be able to make one thread faster. If it could be done in parallel threads, it would all be more efficient. With a date timestamp I'm limited to a single thread, those are the limitations in the threading model of the application. The threading model creates bottlenecks all over the application.
