Sorry, I always get this wrong: the selected rows are extracted first, prior to the randomisation being applied, which is why you are getting a variable order but from a constant selection of rows. So instead of specifying rows 0:5, which will always be the first ones, we need to provide a list of 5 random, potentially non-contiguous indices from across the whole span of row numbers. So if there are 10 rows in your conditions file, you could try an expression like this in the selected rows field:
choice(10, size=5, replace=False)
In the “begin experiment” tab, you would first need to import that function:
from numpy.random import choice