Superposition
Several things stored in one vector at once, all still readable.
Like you're twelve
A window at dusk shows you the garden outside and your own room reflected back, overlaid in the same pane. Both pictures are fully there. Your eye can attend to either one. Nothing was averaged, and nothing was filed in a separate place.
That is what a superposed vector is: several patterns occupying the same storage, each recoverable, none of them in a slot of its own.
Precisely
Superposition is the state produced by bundling. It works because high-dimensional vectors are near-orthogonal: each term contributes almost nothing along the others' directions, so a query aimed at one term is barely disturbed by the rest.
The trade is graceful degradation rather than a hard limit. As terms accumulate, similarity to each individual term falls and the interference floor rises, until a query can no longer tell a stored term from noise. Nothing breaks at that point; results simply get less reliable — which is why fidelity is reported rather than assumed, and why a Cleanup memory earns its place.