On the forecast time aka TTC, it's an inherent issue as there are different lengths of tasks circulated. I doubt that WCG, although there's a reasonable predictability that T000/T001 are the shorts and T300+ are long, splits the fpops header filling and goes with the total app stream average feedback loop. Since there was a pile-up of T000/T001 in the validator, which got force revalidated, that fpops value will have been pushed down too much.

Who the many complainers are I don't know, but I'm not one of them, since the BOINC mechanism is simply not fit for dealing with substantial duration mixing within a single science [with the well known credit #$%@& side-effect to go with it, thusly accept what it is [with a small population at that].