4.7-5
Part (b) has a nasty example, because it's impossible to find an exact value for p, it's 1<p<2. Then, the integral required in the Akra-Bazzi method is \int\frac{dx}{x^{p-1}\lg x} which is an exponential integral, equal to \ln2\Ei((2-p)\ln x), and the solution to the recurrence is T(n)=\Theta(x^p\Ei((2-p)\ln x)).
My guess is that it was unintended, and I believe that slightly modifying a_i, b_i, or the function f(n) can result in an easier recurrence to solve (even without knowing the exact p). I'm going to contact the Authors about that.
Authors contacted about the difficulties in the exercise. There hasn't been any update on this since 3 months now.