Thanks a lot for your contribution. I am sorry to hear that you feel like there are open issues to be addressed in the jackrabbit, sling and felix projects.
I am more than happy to point you the respective documentation on how to precompile the JSPs or run CRX on top of a relational database as the persistence tier and your other more specific comments. Please feel free to contact me if you are interested in details in these areas and I am convinced that we can correct your perception.
Since you mention specifically our open source projects sling, jackrabbit and felix (and OSGi in general) with regards to your perception of instability we would definitely be interested in a greater level of detail of your specific use-case since we found them to be very stable in production environments for years. As you know we are developing all of the projects you mention at the apache software foundation  so please feel free to file issues in the respective JIRA tracker or even file a patch if you have a specific issue to report or even a recommendation for improvement.
thanks again for your comment and please keep us posted via the respective open source issue trackers and simply via our support.
You don't have to feel sorry.
I'm sorry that you feel sorry to people who dislike your products.
Could you tell me how to precompile JSP files and deploy? And, I would like to turn off JSP compilation functionality on live servers completely.
What's the benefit of using database pm? I'm aware of tarpm bug where it deletes data tar files during tarpm optimization and index merge. Does databasepm have the same problem?
If I use databasepm, can I be free from regular maintainance such as datastore garbage collection, tarpm optimization..etc? I find it very difficult to run such maintainance scripts on a live server where the repository is heavily utilized 24/7.
I wouldn't mind shutting down the server and run those maintanence offline. Actually, it would be much better if I could run through maintanance offline.
Also, I heard that there is a bug for share nothing clustering for CQ5.4. I'm not sure if this is jackrabbit bug. If CQ instances are all clustered and I can take one out at a time for offline maintance, it would be superb.
And, what about backups and restoration? Why aren't there set of maintenance scripts? There are documentations with script snippets on knowledgebase... But why aren't there official maintenance tools?
And, where is a book? There are documentations that explain what dispatcher is, what repository is, how to develop custom components, how to configure repositories, ... etc. But why isn't there a book that covers architecture for production site, application development practices, maintenance know-hows?
Does this all mean that Adobe does not care for CQ, and it will not pursue CQ as a viable technology stack for ADEP WEM? (which will be an excellent news by the way).
Or is it Adobe's scheme to monetize support (less publicaly available documents/books.. more support tickets)?