Okay so heres my 5cents about it. Every single project for interlude with the exception of acis is a pvp server project because developers are not skilled enough for real changes, only for custom things that can be found on the net.
As for the base and changes you intend to implement:
L2j based packs or l2j itself are near impossible to make good, you simply need to devote months and months work just to make it "okay".
The augmentation system needs to be reworked completely and that brings the structural changes so thats okay and not even a hard task.
If you make a new scriptengine i suggest you think about making the script cache work because its a really great performance improvement but doesnt work on l2j.
The l2j mmocore always was and still is just bad. Yet i dont think you will be able to adapt any decent mmocore because of how hard it is.
Retail drop protection is not the one l2j currently have so you cant sync that.
Remember to archid and the annotation config^^
When you split the ls and gs make sure to implement the commons too(i suggest you use maven for that).
The geodata and pathfinding of l2j is slow.
In general the use l2j makes this project harder than it should also using interlude makes no sense.
Also i'd like to offer you a position in my pack, i know its not interlude but im pretty sure we can figure and a way to make it benefitial for the both of us, Or even to add your project on my svn so i help you with your interlude and you help me with gracia final. You have my skype so if you are interested we can talk about it anytime.