the key fact is that pi(x) is irreducible. in particular, pi(x) is NOT a unit, and so deg(pi(x)) > 0.
(irreducible elements are non-units by definition: we want to eliminate "trivial factorizations" like u = (u2)(u-1), or even worse: u = (-1)(-1)(u)).
pi(x) clearly divides the "whole product" (that is, f(x)), and what we really want to show is that it divides "the q-part".
so consider (1/v)(f(x)). we have:
(1/v)f(x) = (u/v)p1(x)....ps(x) = (pi(x))((u/v)[p1(x)...pi-1(x)pi+1(x)...ps(x)])
which shows (1/v)f(x) is a product of pi(x) and some other polynomial in F[x] (namely: (u/v)[p1(x)...pi-1(x)pi+1(x)...ps(x)]).
but: (1/v)f(x) = (1/v)(vq1(x).....qr(x)) = (v/v)(q1(x)...qr(x)) = q1(x)...qr(x),
so pi(x) must divide this polynomial as well.
then we apply theorem 8.4.5 to show that pi(x) divides some qj(x) (theorem 8.4.5 is KEY; it proves "irreducibles are prime" in F[x]).
this shows that F[x] is a unique factorization domain....actually we didn't even need "the polynomial-ness" (if that's a word) of F[x], the ONLY thing we need is the "euclidean function" d (deg(f) in this case), so that we have a division algorithm to find the gcd of 2 elements of F[x]: ALL euclidean domains are unique factorization domains, the proof is pretty much the same, and in fact, is the SAME proof used to show integers have a unique (up to sign, and order of prime factors) factorization into prime numbers (which are precisely the "irreducibles" of the ring Z).
note we used the fact that F is a field in one crucial step...asserting that 1/v exists. it is important that neither u nor v be 0 (the units of F[x], though, ARE the units of F, that is: the non-zero elements of F).