I disagree with increasing your cache by so much since that means the WUs are not returned in a reasonable amount of time (especially if you are a wingman) and your system will fall out of the reliable device category.

I'm not sure, but the 7 day (B Type) and 8 day (A Type) deadline tasks (only for reliable, short return devices) will come for a little and then probably dry up leaving only the C types... well they will be 99% of the work, so who worries.

Then, there's the weighting in the feeders... will that remain same?... will that lead to more quasi DDoS type hitting on the servers? Got my thoughts on that, regardless if the WCG hosting service is robust enough to withstand that or not. When we came out of the outage last week the techs manually switched on elements not to have the house topple again.

Mr downtowndan, if you have the urge to cache up, do it slowly... increment in little steps and your client will learn, else you get what KerSamson warns for... there's 22 million tasks coming and I understand it will be in a steady kind of stream when that rain comes... get your raft out.

All right, I did'nt read attentively ALL the posts, but, as you can see, I'm not a native english speaker and reader, and sometimes, I miss some subtleties... (subtilities...?)
Another point is that I don't watch WCG all day long. I agree with giving my computer time, I try to distribute it as equally as possible, I try to look at forums as far as I can, and that's all.
During most of the time, I just need a system that works alone and compute for all projects. That's what happened before DDDT2, more or less.
I'm just surprised that it needs so much watching time and complicated manipulations to participate to this project, that's all.
In fact I don't really mind. The project is running, the scientists wille get their results, and I do crunch on easier access projects, and I hope as useful.
Thanks to all for answers.

Well, this thread wasn't a waste. It gave me some ideas!