I don't have any answers regarding the sampling rates, just some observations:
I did an extensive test with a low frequency, low distortion sine wave and the ~10 Hz sample rate is real. The values obtained are accurate to the expected values for that signal, to the resolution and accuracy of the meter.
So, all well and good, it's a nice feature when you are connected to a computer.
And then there is the update rate of the meter itself; 2 per second. Some people have a problem with that, but I'm OK with it, particularly with the fast bar graph. Fast changing numerical values and be difficult to interpret. A bit faster, or selectable would have been cool.
I did some reading about the A/D and multimeter subsystem chip and found that the conversion rate is 2/ second, so how they are getting 10/second I really don't know. It's been a while, but I wonder if the meter chip does averaging internally and it can be enabled/disabled.
Alternately, it wouldn't be the first time for a manufacturer to push a chip above spec, but 5x is a big jump.
One thing I noticed, and mentioned before, is that if the continuous transmission is selected (10 Hz), the MAIN data tracks the input signal at that rate and the BAR data is as fast, or likely faster than the 9600 baud transmission allows.
However, if you select the single reading mode, and run it in a tight loop, the BAR value changes very quickly, but the MAIN value update rate tracks the big display on the meter (2/sec)
I would be interested to see what you find out about the test mode. I know nothing about the Android app, but I am guessing/assuming that it is Java. Have you decompiled it with something like baksmali?