The HCC tasks are too short, and this messes with Boinc estimations for mixed project setups. If they were grouped into bundles of say 8, or some technically or scientifically meaningful number, it would reduce such overhead. Other projects use this method.

The dddt2 situation has changed many times wrt task allocation.
Has anyone considered the idea of bundling these tasks with other projects; you have to do two other tasks to get one dddt2 task?
As for feasibility, no idea and it is a bit draconian, but might ease the situation.

Doing manual updates is a waste of Internet bandwidth and time for the users as well. I usually just leave the project ticked for all systems, but I sometimes ask for tasks to push the boat along a bit. I probably got well over half my tasks by asking, repeatedly. When I get a few I usually don't bother for a while (weeks) and just go back to several projects for each system (and only get the odd dddt2 tasks, a handful per week if I'm lucky).
Don't know how many hours would be spent WCG-wide doing manual updates, but if you added it up since the start of the project and multiplied it by all the crunchers who use this method, it's bound to be months if not years. Making science a waste of time is surely undesirable.