4.37 is scheduling based on first report deadline.
As other projects seem to provide new work one (Einstein) or 2 (LHC) weeks in advance of report deadline, this means that Climateprediction never gets any CPU time, as the report deadline is normally one year from download.
Don't think this is fair and needs to be corrected.

HStruik

---

There's a very long thread about this on the number crunching section of the SETI board which I gave up trying to follow. But the scheduler is certainly not intended to have this effect, and I don't think we need fear a conspiracy. The issue is complicated at the moment by the very erratic recent history of LHC in particular in issuing WUs, which means that project is 'owed' quite a lot of crunching time by participating PCs.

I suspect that there will be yet further tweaks of the software before we get to the next release version.

The dangers of BOINC creating the situation referred to here were debated at some length on the CPDN community forum about a year ago and so are not a surprise. Trouble is, so many variants have been tried and all have drawbacks. For example, ignoring deadlines altogether isn't right either.

---

The new scheduling algorithm works closer what what we're used to (round robin project switching) with small caches. Try setting your 'connect to' preference to a day or less and run for a few days and see what happens. You might also need to work through your existing cache before you'll see a change.

On my system, BOINC switches to CPDN regularly as it should.