Spin a coin and it ends in a rising shudder that climbs to a buzz and stops dead. Everyone hears it as the coin spinning faster. It isn't. Two rates are running in opposite directions, and the one you hear is not the one you can see.
Watch the two numbers. The green one — how fast the disc itself is turning — falls toward zero. The red one — how fast the contact point races around the rim — climbs without limit. The sound is the red one.
The geometry is solid: for a disc of radius a rolling without slipping at a small tilt α, the contact point circulates at Ω ≈ √(4g / aα), so Ω ∝ α−1/2 — as the tilt collapses, the rate diverges. What α does in time depends on where the energy goes, and that is the part still argued about: Moffatt (2000) derived α ∝ (t₀−t)1/3 from viscosity in the thin air film, giving Ω ∝ (t₀−t)−1/6; a refined air-boundary model gives −2/9. Toggle them: over most of the run you cannot tell them apart, which is exactly why the argument lasted. Real discs are observed buzzing at roughly 20–70 Hz, some past 100 Hz, then stopping abruptly — the yellow line is roughly where something else takes over and the model stops being about the world.