How large must $y(\varepsilon, n)$ be so that every interval $(x, x+y)$ contains at most $\varepsilon y$ integers having a divisor in $(n, 2n)$? The candidate proof gives the sharp fixed-$\varepsilon$ order $y = \Theta_\varepsilon(n)$, uniformly in the translate.

How large must $y=y(\epsilon,n)$ be such that the number of integers in $(x,x+y)$ with a divisor in $(n,2n)$ is at most $\epsilon y$?

A linear scale is known to suffice (see erdos_450.linear_scale_suffices). Whether the optimal scale is *sublinear* — a sufficient Y with Y ε n = o(n) — is open.