The Hierarchical If-and-only-iff problem (Watson, Hornby & Pollack, 1998) is the deliberate opposite of the weasel: a fitness function built from nested building blocks that conflict at every scale. A block of bits is rewarded only when it's all-0 or all-1; blocks combine into bigger blocks, all the way up. Mutation and hill-climbing get trapped; crossover is what assembles the hierarchy. Watch it build, then Race mutation against recombination.
Each row is a level of the hierarchy: the bottom row is the raw bits, and each row above merges pairs from below. A cell lights up only when its whole block is uniform. Watch order assemble from the bottom up, and notice how small blue and pink blocks have to agree before a larger block can form. That agreement is exactly what mutation struggles to negotiate.
Both lanes use the same algorithm (deterministic crowding, which preserves population diversity), so the only difference is whether children are made by recombining two parents or just mutating one. Watch mutation's fitness plateau on a patchwork of conflicting blocks, while crossover climbs in steps, each step a level of the hierarchy snapping into place. This is the building-block hypothesis made visible, and the heart of why evolution invented sex.
Watch the mutation-only lane freeze with a fitness short of the top. It has solved every small block, but
some as 00 and some as 11, which can never merge. Escaping needs a
coordinated multi-bit flip that single mutations won't make.
All-0s and all-1s are both perfect, and they differ in every single bit. Runs settle on the all-blue or all-pink pyramid at random; the symmetry is exactly what makes the top of the hierarchy so hard.
Drop the population too low and even crossover stalls. Recombination can only combine building blocks that actually exist in different members, so it needs enough diversity to draw on.