Yeah, for actually storing transactions the approach Satoshi uses of
relying on a database engine makes sense and is what the code already does,
so I'm not sure why this is a problem.

The real problem with Satoshis code for scaling down to smaller devices
(and one day desktops too) is the need to store all the chain headers in
RAM. BitcoinJ avoids this but just creates more problems for itself in
other places, partly because we also try to avoid a database engine
(read/write traffic on phones can be insanely expensive, especially on
older ones, and so sqlite is known to be a serious cause of performance
pain on android apps).