Here's the "not sure if it works"

the thing is about checking WU's the thing is that statistically you don't have to check all incoming WU's you need to take a part for analysis and then produce some kind of report and take soe decision.
So in fact there's a lot of CPU's that works well and there's no need to check all their WU's you just need to take random number of it to recheck it - this is a huge data that will not be analyzed twice and furthermore (what's that word means?) you can check only the part of WU analysis so it will take less time, kinda a random part will be taken and marked inside the analysis these marks will be posted in the "need to check" job and these parts will be then parsed (i'm not really sure maybe it is working already so just giving ideas)

Actually i've started a theme for new ideas here
http://setiathome.berkeley.edu/forum_thread.php?id=69850
here's the part of message

But actually this bug produced a new thought for me - maybe this is the way of getting attention for those jobs? and i've imagined that for broken and invalid circuits actually should be produced more jobs not only ot the computer that made the mistake but also to recheck it not just 1 time but 2 or 3 times on other computers and of course there should be a simple button with process dialog that will restart the job, but also will save the invalid job to produce an analysis of error, so this should be descripted by author and maybe some other volunteer that has the opportunity of checking the broken ring (loop, job - nvm).
Also i've suggested a new function in my own thread that will get the opportunity of visual look of already processed signals in time from universe specialized coordinates, so for example i can open alpha way (alpha-centauri) and see all signals that have been already done, then i can predict myself the time and the cluster of stars that i would like to analyze and that is not checked yet, and then i'll start those jobs from different satellites that will get me the information from that particular cluster, so even people can be organized by those star clusters and will try to get signals by activity that they feel they want to parse so this is mostly not checking of all signals but it's about a human nature and feeling.
and the program should take some signal prediction a give some coordinate moving in time probably or people can take it manually too.
well this is about not to try to analyze the whole white light but leave it for humans to decide about time and space.

it's pretty easy
i'd like to see a graph of whole pattern downloaded in my personal view and knowing then i want to pick manually some parts of pattern that i think can be interesting. even more i'd like to pick those patterns from the main channel so i can browse signals on my own and simply crop or select time and satellite, see the picture of signal select the part and try to analyze it, also i'd like to know a history of the particular space point, kinda i got signal from alpha and i need only signals that goes from that part of the universe in time, well this should be not less than a month as i think or at least a week.
so to catch the idea you of course understand that satellite is like a microphone and depending on it's software this can listen any part around the earth orbit or can be focused, so i'd like to have a focused and satellite indepent signals from some part of the universe and actually i'd like to have an option to reorganize those signals in time.
this is like a timeline probably. i can explain more if it still not clear.
Well for me it is very interesting to try find signals in the way of the vector that the solar system is moving forward, just to know if the road is safe you know... and also i don't like to analyze signals that i suppose will not be interesting, since i can think - i can try the way of not the brutforce analysis of everything but of my mind one, that can help me more especially in life finding since it's the basic function of all living beings.
Well about the errors now - the thing of error is that it is like a random generator - so statistically you should take a random part of your issue to determine if all goes well and your script is working true, when you get each 10th part of issue - it is not the random part it is what you actually programmed but whorse it is a math computation dependent on a lot of human and computer work - so it is not the real random? while the real error - is real random.
the other side is the "happen" it's happened so why not to really get some attention on that case and another one is that errors can be produced by reality to specify the moment and one more is that actually error can be something that computer cannot actually understand so it is something not defined by the rules and this is what you can interpret as a signal that you didn't defined really - well mostly because of your human nature, but also because of your human nature you should try to interpret the information by yourself and the result can be more significant because the real consciousness knows about your human nature.

I can see your suggestions working well in a more perfect environment.
However due to the massive differences in volunteer's computers and various platforms,
it would not be possible to get scientifically accurate representative data from a single processing, at least not as the system works at present.

IMO ..Part two of your posting sounds like you should consider organising your own boinc project.
What your proposing would not cooperate with the agenda of the project scientists here. And, it would essentially require an entire restart from scratch.
And, a much more devoted radio observatory than we presently piggyback our data from as cheaply as possible.
I don't think BOINC would even support such a user involved application,
You would probably have to develop an entirely new infrastructure to get your data processed.