Yes, it is a new constraint, but it does not follow from Maxwell's equations. Like any gauge condition, it's an optional constraint, which you can choose to obey if it makes your life easier, and disobey if it doesn't.
In a way, it's a tradeoff. You can choose to take the Lorenz gauge; if you do that, the equations for the fields will become simpler, but you have to make sure that the fields satisfy the gauge condition. Or, you can choose not to bother with gauge conditions, but you'll have to use more complicated equations.
You seem concerned that the gauge condition should already appear in the solutions to the equations. But an important fact that is not often discussed is that a solution of the most general equations for the potentials will have arbitrary, undetermined functions. This is exactly analogous to the arbitrary constant we find in the electrostatic potential: to fix it, we have to specify the value of the potential at some point, which is our gauge condition. The constant is completely unphysical, so we can choose it at will. Or we can not choose it, which is fine too, we'll just have to drag it around everywhere.
In fact, the analogy is stronger: the fundamental, non-gauge-fixed equation in electrostatics is $\nabla^2 V = -\rho$. The general solution for this is
$$V(\mathbf{x}) = \int d^3\mathbf{x}' \frac{\rho(\mathbf{x}')}{|\mathbf{x}-\mathbf{x}'|} + f(t),$$
where $f$ is an arbitrary function of time. You cannot derive $f$ from the equation, it's undetermined - and of course, this is not a problem, since any $f$ will give the same field. What you can do (which is what we do implicitly all the time) is use a gauge condition $\partial V/\partial t = 0$, which will force $f$ to be a constant. Again, this condition doesn't come from the equations: you choose it yourself, to remove some of the arbitrariness in the solution.