The "Holy Code" problem
The problem that I've seen with OpenOffice is the same issue that happened with OpenSSL. You have a core group of outstanding developers build a specific product; product then becomes hugely successful; original architects and developers leave/retire; technology undergoes many major shifts; original code is left in since its assumed that because ti was brilliant code initially, its brilliant code now.
This is why OpenSSL retained its own malloc-like function, no one wanted to break the critical code even though the reasons for it existing are no longer applicable (The Standard Libraries of many OSes at the time lacked a standardized malloc so the solution was to either bake a malloc into the code or write piles and piles of OS-specific code).
With OpenOffice, these issues rear their ugly heads when it comes to drawing the UI (Since TK / gTK hadn't matured by the time OOo was written) and for working with the OS (just like OpenSSL, OOo was written before many of these things were standardized across platforms).
At this point, the only thing that could really save OOo is to do a feature freeze for now and focus solely on cleaning the code that's already in there as well as cut stuff that is needed to run on older OSes / architectures and let the OS handle a lot of the grunt work.
Biting the hand that feeds IT
