Refining multiple datasets simultaneously on the screen can look a lot slower that if you each data pattern one at a time. It looks slower as it is slower. It takes longer because if you are refining N parameters on the screen, the matrix inversion is a N^3 (cubed) routine. Thus if you are refining 40 files with just one peak in each file, (1 peak has four parameters; position, area, width and shape); that is 200 parameters in one refinement.
One quick trick if the time is taking too long because of the matrix inversion is to open one window at a time, and cycle through each file using the ~ key after you finish each individual refinement.
Another via the people at the Royal Institution in London to get around this problem; is by running multiple instances of XFIT; with each instance of XFIT loading roughly 6 data files (depending on the complexity of the peak fitting to be implimented).
Multiple files in a project file being refined simultaneously.
Multiple files in a project file being refined one at a time and using the ~ key to cycle through each one.