Finally found benchmarking bug that would sometimes cause invalid benchmark data to appear. This would happen if the DF Client was writing to progress.txt at the exact same time dfGUI was trying to read from it getting partially written data (that is why I couldn't reproduce it). I have now put in a check for this and it should no longer happen. Special thanks to Joe O. and everyone else for helping me track down this bug.
Fixed a couple of minor bugs I found while searching for the benchmarking problem.
Added support for new Use RAM feature in DF Client that gives you twice the speed if you let it load the entire protein.trj file into memory (-rt & -rf). A 200AA protein would need 143MB RAM, a 150AA protein about 121MB and a 100AA protein about 89MB of RAM.
Fixed Upload bug where dfGUI would not always wait for the DF Client to finish uploading before it tried to restart DF Client in regular mode.
Comment