Turn-based 4X is honestly a better host for this than a racer, because Civ strategy is secretly all about global structure — encirclement, chokepoints, "safe" back ranks, circumnavigation — and global structure is exactly what topology governs. The racer only used topology locally (seams, wrapping); a 4X uses it strategically. Some of these are gimmicks and some are genuinely deep, and interestingly the flashy ones are the shallow math and the invisible ones are the deep math. Going system by system:

The tile grid already forces your hand (Euler characteristic). A pure hex grid, three hexes per vertex, gives χ = F − E + V = F − 3F + 2F = 0. So an all-hex world is necessarily a torus or Klein bottle — this is why Civ's cylinder/donut maps feel natural and why every "Civ on a sphere" attempt sprouts exactly 12 pentagons (χ = 2, and the defect count is 6χ for 3-valent tilings). Flip the sign: a genus-2 world needs 12 heptagons. Design gift: don't hide the defect tiles, make them the map's 12 sacred sites — monoliths, natural wonders, Planet's nerve nodes in SMAC terms. The combinatorics hands you your landmark placement for free, and players will feel that these tiles are "where the world is pinched" without knowing why.

Walls and borders: homology is the real strategic layer. On a sphere, any closed wall encloses something (Jordan curve theorem — which Lee actually proves, via Seifert–Van Kampen, I'm fairly sure it's in Ch. 10). On a torus, a Great Wall built along a nonseparating cycle encloses nothing — both sides are the same side. So "did my fortification line actually partition the map" becomes a question about whether your wall's class in H₁ is zero. On genus-g you have 2g independent nonseparating directions, meaning defensive play has 2g distinct ways to be wrong. Same math governs naval blockades: between two cities on a torus there are infinitely many homotopy classes of trade route, and camping one strait only kills the classes through it. Pirates must cover a cut system, not a point. This is the deepest fun in the whole idea and it's completely invisible — no mirror units, no spectacle, just veteran players slowly realizing their intuitions about "enclosed" are theorems that stopped being true.

Exploration = the player literally constructs an atlas. This is the SMAC-flavored one. You start not knowing what world you're on. Fog of war means every explored region looks like boring ℝ² (that's the definition of a manifold — locally you can't tell). The genus only reveals itself through loops: a scout circumnavigates and either does or doesn't come home, does or doesn't come home mirrored. Make topology-detection into tech: mid-game "Planetary Cartography" computes H₁ of your explored territory; the satellite endgame reveal is literally the classification theorem announcing "you are on the connected sum of 3 tori." Non-orientability as a jump-scare: your first circumnavigating former returns with its terraforming kit on the wrong side. On RP², since π₁ = ℤ/2, circumnavigating twice is homotopically trivial — Magellan's achievement can be undone, which is a very SMAC koan.

Topological terraforming (the late-game bomb). Civ2/SMAC already let you raise and lower land — the natural escalation is engineering that changes χ. A "wormhole handle" project glues a tube between two distant tiles. This is connected sum, straight from Ch. 6, and its strategic payload is nasty: your opponent's separating defensive wall can become nonseparating the moment a handle appears near it. Their border still exists; it just stopped meaning anything. Retaliation is closing handles (surgery), stranding trade routes whose homotopy class no longer exists. The planet's H₁ becomes a contested resource.

The weirder "topological spaces, not manifolds" tier — since you asked. Wedge two spheres at a point: two planets touching at a single non-manifold pass, detectable in-game because local homology (Ch. 13) at that tile is wrong — flavor it as a survey anomaly, and it's the ultimate chokepoint, a point whose removal disconnects worlds. Non-Hausdorff line-with-two-origins as a Planet-mystery tile that exists as two copies which agree everywhere except at themselves — a unit walking in converges to both; psi-flavored, use sparingly, it's a gimmick but a great one. And the deep-cut shitpost that I'd genuinely prototype: a scenario map on the long line. Infinite west, every local region ordinary, but no sequence of finite-movement turns ever exhausts it — Manifest Destiny as a literal topological impossibility. Compactness, it turns out, is the axiom the entire 4X genre rests on: compact ⇒ finitely many tiles ⇒ scarcity ⇒ conflict ⇒ the game ends. Drop it and you've made a different genre (eternal-frontier expansion sim), which is itself an interesting design fork.

What Lee's book doesn't give you, honestly: distances, unit movement costs, "closest city" — anything metric. Also orbifold-style quotients with fixed points (mirror-symmetric maps where the axis is special) go slightly beyond his free-action treatment. But for a tile game you need less smooth structure than the racer did; the CW-complex chapter is your map data structure, full stop.

If I had to bet on the fun ranking: homology-aware walls/blockades > topology-as-fog-of-war-reveal > handle terraforming > defect tiles > Klein bottle mirroring (fun once) > non-Hausdorff tiles (fun exactly once). Notice the inversion from the racer, where the covering-space rendering trick dominated — here rendering barely matters and invariants are everything, because a 4X is played in the mind's global model of the map, which is precisely the thing algebraic topology formalizes.