optimizing the IO-threads is not that easy (but there are some good ideas out there

, there are several possible bottlenecks. one of the simplest things is to convert straight from the card - it simply prevents the time consuming copy-process to a harddisk - if you summarize all copy/process time, its not that clear, what concept is ideal. i wouldn't focus that much on multithreading but more on one-click-batch-solutions, so you can start your main (editing) work straight after a unattended coffeetime.
and: baldand gave his sourcecode free to help and code, in contrast to me (yet). you're all able to help

regards chmee