We have been pondering over this for quit some time. I guess we can agree on the basic premise that turning off widgets does seem to help (though not fully solve the issue) and certain specific widgets are worse than others. But turning of widgets is not a solution.

I propose that when widgets are promoted from extras-testing to extras they should specify how much cpu\mem usage they consume on an average . This at least would give us some idea as to which widgets can potentially cause slow downs. Or if widgets had there own processes it would have been easier to find out which widget is consuming more.

--Vj