Reverse-Z
The trick of storing depth reversed — the near plane as 1, the far plane as 0 — so a floating-point buffer's precision matches the perspective projection's and distant z-fighting disappears.
The perspective projection crams almost all depth precision up against the camera and leaves only crumbs at the far plane, which is what causes z-fighting in the distance. A floating-point depth buffer has its own skew at the same time: floats pack the most representable values near zero. Lay the two biases on top of each other in the same direction and it only gets worse.
Reverse-Z flips one of them. By mapping the near plane to 1.0 and the far plane to 0.0, the float's dense region lands where the perspective is sparse, and the two very nearly cancel — precision becomes almost uniform through the whole scene. The cost is that the depth test must switch to GREATER and the buffer clears to zero. The trick was known in the 1990s but only caught on broadly in the 2010s, once floating-point depth buffers became standard.