snip
You can regulate any machine either by associating it to 1 of the 4 possible website profiles (default, home, work, school) and adapting these, or if that route is exhausted, working the local device preferences. Principally, those being 24/7 machines, you could set the "connect every..." to 0.0 days, meaning internet is seen by BOINC as always there and the additional buffer to 0.5 days, which will make sure to fetch enough work for 12 hours "per-core°, minimum. Then whatever "Ready to Report" (RtR) is there would be cleared on the next work fetch call... the less important part as the result files themselves have already updated. The RtR only shakes the scheduler and that daemon is the heaviest loaded in the BOINC system, hence the attempt by BOINC to combine work fetches with RtR clearings, one client-server recalculation rather than 2.

If your science selection is for rare work such as DDDT2, then the server will instruct the client to back-off if multiple times no required work was in the feeder. That would though presently not apply as there is a steady trickle of DDDT2, Back-offs for that reason would be at most a few hours i.e. the chance of your computer idling is small with a 12 hour buffer size, unless of course WCG goes off line for longer, so I actually run with a 1 day cache and that long an unscheduled outage has not been seen since X-mas 2005.

--//--

PS, note that the connect interval and cache are additive. If you set a 6 hour connect and a 12 hour cache, you get a total of approximately 18 hours stored on the client. Not for noting it's called "additional buffer"!