There are surely bound to be minute differences in the characteristics of the two sets of pulleys, so a perfectly matched pair of belts would only produce identical degrees of tension if the pulleys and the component mountings had been manufactured to a corresponding degree of precision?
In practice I think there are bound to be small differences, so I suggest that a randomly selected pair of un-matched but nominally identical belts might actually have an advantage - if on setting up you find the tensions are not quite the same, try swapping the belts over.
It's a bit like sliding bits of slate under the leg of a wobbly garden table. Fortunately slates are not made very accurately, so you can usually find a better fit.
Clifford's Theorem - "Perfection is limited in its choices, but errors can often be offset to produce closer results"
