I have the Berkeley app installed on three of my tablets and they all work fine. Android is harder to gain points because of an ARM processor, but still its contributing. I haven't tried the Native Boinc app as of yet, and don't think I will because if it aint broke, dont fix it.

Thank you for the bump lavaflow.
You summarized perfectly why I've chosen Native BOINC. The main reason as I started on Android was that BOINC didn't run on batteryless devices.
I like the other benefits very much too. I control the androids via boinc tasks on my main computer or via webversion or I use andro boinc on my smartphone to connect to my android devices.
I'm using my batteryless android devices for boinc only. That means that boinc configuration is done to reflect this: Full speed ahead for boinc. No limitations for boinc.

My own statistical skills tell me it's very possible, also because the nativeboinc agent is identified as 7.0.36 in the result logs and the bfa versions are 99.9% different, mostly auto upgraded to 7.3.x or 7.4.x. The agent is logged, so please dear technicians, could you dig out the android success / fail rate stats of nativeboinc versus boinc for android. That is, if you still have interest in maintaining support. Few days ago the android returns were the lowest since charting began!

On more good news, not reading manuals, just trying typical android interface things like holding screen on a project to get a pop up menu, it was fairly straight forward to add the tablet to boinctasks on the pc. Used the mac code, as pulled from the router info screens, and entered the usual password for boinc remote and voila, connected. Even connected the tablet to view how the desktop was doing.

The next hurdle was that the bfa agent is set to report results immediately but nativeboinc is not, but how to edit the cc_config.xml in the protected data storage zone? Boinctasks came to the rescue again. It has an edit screen for the config, just as the newest pc boinc agents have, except the bfa agent only allows log flags picking, not options. Thusly done, tasks are now reporting immediately, 'plus', i've set the agent for a delayed start of 180 seconds, as that is one of the sources of task fails. The message log only protested on an unknown vbox_window and fetch_on_update flags. Latter suspected as overriding the 'no need, buffer full', but that's a pure guess.

The most important piece though is the success rate. Preliminary feel, no different than the bfa agent, but then that one already show > 80% success rate opposed to the general > 80 % fail rate for most everyone else. The very first result crashed with a sigill, and one more went with a signal 11, the sensing that the device is too busy is as with the bfa agent latent. 30% it is set for cpu load before pausing. Thanks to the nativeboinc option to tick/untick laim, was able to stop 2 tasks in separate incidents and make them unload after they locked up on cpu time, but not on elapsed. They reset, both to a checkpoint somewhere at 1:30 hours, loosing 10 each, but they were in the android manager eating cpu time, something in the task app not passing this. At any rate, boinctasks was instrumental in showing the red boxes on these tasks showing they had not checkpointed for 10 hours. After their restart, they ran normal to the end. Both went after arrival of wingman to pending verification, 1 invalid after a 3rd opinion and one valid. Is this something the technicians are interested in? Maintaining a detail sheet of outcomes for each task to catcher in the rye.

Summary on the native boinc agent: For those who like buttons and controls, to include remote and have boinctasks log the result history, a must. It's almost the complete pc boinc manager on a small screen.

Summary on results so far:

Processed: 23
Server Aborted: 2 (no reply came in and a task with too many errors taken out)
Valid: 12 (device stats as of noon: Tablet 3125834 08/18/2014 12:03:39 0:005:07:23:44 6,338 12)
Invalid: 1
Pending verification: 2 (candidate invalid or too late due too absence of achievable quorum)
Pending validation: 6 (all ran through without incident)

Since running native, see some quirky notices that the tablet battery is fully charged as if it drains it to below 95%, which is the android os tipping point to show the charging icon and load up again. At least the charger is able to bring it back to 100%, which under bfa it could not on the standard charger. Suspecting this as one a source of the many task fails due the continous stop/resume cycling at the low battery setting point of the agent! On multicores, reduce the allowed processors percent to the point where the charger does keep up to maintain full charge.

Oh and the mentioned first fail was while running on 4 cores. The device ran rather hot so switched down to 75% of processors aka 3 cores. Since it's fine and the set battery temp in the agent at 36% was since not exceeded, continuous computing, when on charger.

Sorry for the essay worth of info waffle (and not interested in anybodies snarkish comment. You are ignored!)

On August 13th I put one of my androids (odroid u3) on wcg again. So far it has returned 35 wu. None has thrown an error or went invalid. One was a resent that returned with too late. Right now I have 15 pending validation and only 1 pending verification.
I'm not sure about the android os version but I think it's either 4.2 or 4.3 in combination with nativeBoinc.

Just my thoughts on this,

I have been using the boinc app (not native boinc) for several months now and have not had any issues with work units having been rejected. Although some work units are a bit long (longest one has been 16 hours, but some are in between 5 and 12) I am happy with it.

When I tried using native boinc I did not like the interface and prefer that simpler one that the non-native boinc app provides.

That is just my opinion.