Calculations for the heat transfer between superheated silicate melt and evenly dispersed 1-mm cold clasts (a model for impact melt rocks) indicate that (1) most of the thermal gradients are smoothed out in 100 s. (2) the extent to which cold debris is melted is slightly more than proportional to the fractional difference of the equilibration temperature between the solidus and the liquidus of the melt, (3) smaller clasts are preferentially dissolved, (4) the melting of clasts depends on the instantaneous local temperature of the melt in the region of the clasts, and hence clasts can melt during cooling whose liquiduses are higher than the equilibration temperature, and (5) the rate of equilibration is sufficiently high that clasts whose dissolution in the melt is slow may be preserved. Calculations for cooling of the 200-m-thick melt sheet at Manicouagan suggest that complete crystallization takes 35 years at 100 m form the edge and 1600 years at the center. Failure to consider the latent heat of fusion results in estimates of these times which are too short by about a factor of 2. The time to cool to a given temperature is proportional to the square of the distance from the boundary of the melt sheet only when the finite nature of the sheet is unimportant.