One of my problems with the BOINC app on Android is that the event log is pretty much useless. Nearly every time I look at it, it only shows setup messages usually timestamped about an hour earlier. In this most recent case, I'm certain the client was running the entire night up until I checked around noon after the task was gone, yet there were no messages about the task in the log, it only showed setup messages timestamped just over an hour earlier. Though, one of those setup messages was warning me of another late task that might not get credit, which has been running at 100% completed for at least 6 hours (inaccurate task progress is another issue I have with the Android app).

I understand what you're saying about the 24 hour thing, and makes sense now that I look at the WU status for my PC results. It seems the reason my PC results stick around so much longer (up to about 2 weeks after completion) is because they require a quorum of at least 2 and at least one other client isn't returning a result in a timely manner, usually missing their deadline. My Android device, on the other hand, tends to be the slow one, so the results are gone within 24 hours of completion if not before completion because it was way late.

While yes, quickly moving the results to the master database makes sense, I would still like a decent opportunity to confirm the final status for my recent returned results and easily spot negative trends (such as certain projects with a deadline often too short to complete in time on the device) without having to check the results page every day. If reading back from the master database isn't a reasonable option, I'm thinking something like a snapshot of the result that was sent to the master database, which could have its own longer expiration or perhaps even go to a separate archive section if it's not too heavy on server resources.

Anyway, thanks for clearing this up. I was almost convinced I was experiencing a bug.

I have a Nexus 5X with 6 cores (4x 1.4GHz + 2x 1.8GHz) and was running 3 tasks at a time. I had it configured to only run with battery over 90% and temp below 40C, even unplugged. I've selected projects that typically have jobs taking less than an hour on my main PC so my phone isn't stuck with really large tasks. The side effect seems to be a much shorter deadline, usually somewhere between 24 hours and 2 weeks. On my phone, most tasks tend to take about 40 hours of processing time to complete. I also noticed that when a task is suspended and resumed, as is very common on the phone, the progress sometimes jumps back quite a bit, from about 20% to 0% in one recent case.

That seems to be the trick. Much faster when running 1 task at a time. That suggests the app may not be utilizing the CPU cores correctly for multiple tasks. This is satisfactory, I don't feel like giving up on it any more.

I think I should have submitted this thread to the Android Support Forum instead of Website Support. Mods, feel free to move it there.

Most puzzling observation, so decided to run my 4 core Nexus, which I always run at 2 for BOINC, as 1 threaded OET1 only and find at 2 threads the mean was 5.04 CPU hours, elapsed 5.13 (98.3% efficiency), and just running 1 thread 5.09 CPU hours, elapsed 5.21 (97.7% efficiency), sample size 10.

Suppose there's apples and oranges amongst ARM CPU's.