So, to back down a little: my main point is that there isn't any consistency between the way *VOs* treat the GOCDB and the BDII's relative importance. ATLAS definitely think the GOCDB is canonical, but LHCb and CMS apparently treat both equally.
This is made worse by the fact that, if you're debugging a system, YAIM tends to automagically set things online in the BDII, and sysadmins tend to assume that they're protected from this breaking jobs because the GOCDB is always set to have things in Downtime.
There's at the very least an education problem here: in the Ops Team meeting, there has been quite a lot of confusion raised over what precisely is important to set at what times when services are Down or not in Production, and many people seem to have been surprised by the "official" answers to these questions.
It would be nice if we had, at least, some flag for YAIM to tell it "yes, this site is in downtime, please don't ruin everything", I suppose... or if we could get the VOs to be consistent as to the precedence they gave their various information services. I don't suppose there's any way to pressure them?
Sam