Logging Memory Space

I am using a number of RPros to capture data where the synchronization among devices is critical. Assuming that their internal clocks cannot be synchronized I tried to synchronize them after the fact by shaking them in a specific way at the same time. But it turns out that they don't all record it the same way when they are all in my hand. The accelaration amplitudes vary so much, and the timestamps vary so much, that I have to wonder if the logging memory filled up.

According to your documentation, there is only 10k of logging memory. Therefore, if I am logging at 50 hz, I should run out of space in about 50 seconds.

Yet some of the signals appear to last nearly two minutes.

Also, on these forums in some places you indicate that logging data should be used when timestamps are critical. Why are mine so far off? To the point where they don't even seem to have covered more than a few seconds of the same time. Are they devices turning off after a minute?