Okay, here's what I'm planning to commit (tomorrow), in case anyone wants to give it one last check.  I've added some comments half-explaining the algorithm used (just in case the referenced URL goes out of existence).  I made a couple of other minor changes to the algorithm:

(1) in factorial_partial_product, use 'k = (n + num_operands) | 1;' rather than 'k = (n + num_operands - 1) | 1;'.  This saves an operation, and means that when a range including an odd number of terms is split then the bottom half gets the extra term, which makes the partial products a teensy bit more balanced, since the bottom half consists of smaller numbers.

(2) in factorial_loop, I set the initial value of 'upper' to 3 rather than 1.  This avoids factorial_partial_product ever being called with a start of 1.

Apart from that, no significant changes.