The OP could be asking either how fast the singularities hit or how fast the event horizons hit.
The singularities are spacelike, so it's not well defined to ask about their velocity. The velocity of A relative to B is only something we expect to be well defined if A and B are timelike or lightlike world lines (or surfaces with similar properties). A singularity is basically the end of the world for observers inside the horizon, so asking how fast one singularity is moving relative to another is like asking how fast the world is moving relative to someone else's end of the world. It doesn't make sense. The end of the world is a time, not an object.
GR doesn't have global frames of reference or a way to measure the velocity of an object relative to a distant observer. See
How do frames of reference work in general relativity, and are they described by coordinate systems? . So there is not any meaningful sense in which we can talk about the velocity of an event horizon as seen by some distant observer, only by an observer near the horizon. So this means that if we were hoping to get an answer about the speed at which the event horizons collide as they merge, it would have to be the speed as measured by an observer right near the point of merging.
An event horizon is a lightlike surface, so in general all observers near a horizon say that a horizon is moving at the speed of light. So the observer probably sees both horizons as moving at $c$ as they collide. This doesn't mean that one horizon is moving at $2c$ relative to the other horizon, because we don't have frames of reference moving at $c$, i.e., there can't be an observer at rest relative to one of the horizons.