Multisample load times (technical FR, not sure if it helps)

I’m not sure if this would help, but probably it would.

So NAND flash SSDs are really fast, but they only live up to their full performance potential (forget maximum read / write MB/s, IOPS is whats important here) when the storage interface queue is full (“QD”). This can be accomplished by using multiple threads to load data (lots - some SSDs only saturate at a QD of 32, meaning 32 threads for loading files).

Maybe this would be a nice idea for faster program loading?