I do not know much about the relation between sedimentary layers and fossils contained therein beyond common sense and what's been posted in this thread. But from what's been posted, this doesn't look like circular logic to me. It looks like an unconstrained optimization problem, which on the surface, yes, looks like circular logic, but is not.
I deal with these types of problems a lot in engineering. You know there is a relationship between variable A & B through several other variables. But you don't know what the values of A & B are, or rarely any of the other variables for that matter. And unfortunately, your job is to find out what the best values of A & B are. But you don't have a reference point to start from, so A & B can be an infinite number of values.
I have to go to work and don't have time now to extrapolate on the math, but basically there are many valid ways to find the values of A & B. (The oldest probably being the golden-section search, and newton's method) They all involve initial guesses, and using that initial guess to recalculate the initial guess generally with an independent variable(s). Through several iterations, the calculated values for A & B will converge close to the real values of A & B.
On the surface that looks like circular logic, since you used A to find B and B to find A, when you didn't know the true values for either. But it's not. It's using a mathematical relationship to identify maximal/minimal values which correspond to the true values you're looking for. In this case, I would infer the key/index fossils are those which are put into the unconstrained optimization equations. By using multiple fossils, the convergence of calculated values would be even closer to the true values.