The example you gave is a degenerate case that isn't really representative of the real world situation. Making x to be an odd multiple of pi in your example is an excessive constraint on the problem. Real world problems requiring creativity exist in highly multidimensional spaces and their fitness functions often exhibit fractal behavior.
For example, let P be the set of all English-language poems of N lines, where each line is a semantically and syntactically valid expression in English and N>0. P is an incredibly huge set. We want to find solutions of high aesthetic fitness on the set P. Applying a few arbitrary constraints reduces the size of the search space but has little effect on our ability to find solutions with high aesthetic fitness. For example, we could arbitrarily restrict our search to the space of English poems with N=14, 10 syllables per line, an alternating pattern of an unstressed and stressed syllables in each line (iambic pentameter), and a rhyme pattern a-b-a-b-c-d-c-d-e-f-e-f-g-g in the terminal words of each lines. We've now drastically reduced the size of the search space such that we are now searching only the space of valid Shakespearean sonnets. However, even with our arbitrary constraints in place, we're still solving a highly unconstrained optimization problem and the arbitrary constraints we applied don't prevent us from finding solutions of high fitness.
We could say that the bumpiness of the fitness function is fractal -- the shape of the fitness function on an arbitrary subspace resembles the shape of the fitness function taken over the whole space, providing that the subspace is itself "large" in some sense. The point is that there's nothing special about the arbitrary constraints we imposed. We could instead choose to work in the space where N=16, the number of words per line is between 4 and 7, the first four lines are all questions, and the last line is constructed in the form of a direct object followed by a subject and a verb and a repetition of the direct object (e.g. "Yoda, you seek Yoda"). There are still many aesthetic and unaesthetic solutions in this space, perhaps there are even some solutions with greater fitness than any of the 154 solutions that Shakespeare discovered with his set of arbitrary constraints.
For example, let P be the set of all English-language poems of N lines, where each line is a semantically and syntactically valid expression in English and N>0. P is an incredibly huge set. We want to find solutions of high aesthetic fitness on the set P. Applying a few arbitrary constraints reduces the size of the search space but has little effect on our ability to find solutions with high aesthetic fitness. For example, we could arbitrarily restrict our search to the space of English poems with N=14, 10 syllables per line, an alternating pattern of an unstressed and stressed syllables in each line (iambic pentameter), and a rhyme pattern a-b-a-b-c-d-c-d-e-f-e-f-g-g in the terminal words of each lines. We've now drastically reduced the size of the search space such that we are now searching only the space of valid Shakespearean sonnets. However, even with our arbitrary constraints in place, we're still solving a highly unconstrained optimization problem and the arbitrary constraints we applied don't prevent us from finding solutions of high fitness.
We could say that the bumpiness of the fitness function is fractal -- the shape of the fitness function on an arbitrary subspace resembles the shape of the fitness function taken over the whole space, providing that the subspace is itself "large" in some sense. The point is that there's nothing special about the arbitrary constraints we imposed. We could instead choose to work in the space where N=16, the number of words per line is between 4 and 7, the first four lines are all questions, and the last line is constructed in the form of a direct object followed by a subject and a verb and a repetition of the direct object (e.g. "Yoda, you seek Yoda"). There are still many aesthetic and unaesthetic solutions in this space, perhaps there are even some solutions with greater fitness than any of the 154 solutions that Shakespeare discovered with his set of arbitrary constraints.