Each language keeps a category as a cloud in acoustic space. Projected onto the axis that best tells them apart, they become two Gaussians: the closer the centres, the more they blur. The dashed curves are speakers of one language only; the filled ones are bilinguals — who tend to compress the gap and live in the middle. Push the centres apart or tighten the compression and watch the ambiguous zone breathe.
There's no point where two categories “become one”. There's only a curve: overlap grows as separation shrinks. The blue dot is the monolingual speaker; the amber one, the bilingual — always a little higher up, in a slightly more shared space.