Villermaux and Bossa reply:

Collisions between drops, when enforced by an artificially confined channel (as opposed to a free turbulent air jet with turbulence levels much more representative of the medium at a cumulus base than those achieved within a potential steam, carefully prepared in a laboratory wind tunnel) may influence their size distribution. It is also true that collisions do occur in real precipitations, and can mediate a splash-like interaction. When it actually occurs, the phenomenon is very similar to bag break-up1: the formation of a corrugated toroidal rim (disk), suffering a capillary instability producing thinner drops; their size distribution is likely to be similar. That scenario would also be equally consistent, as ours is, with the observation that apparent fall velocities can be larger than the terminal velocity for a given size2. Fragments have, transitorily, the same velocity as the mother drop(s) before relaxing towards their equilibrium fall velocity.

But random collisions are, at best, a two-particle process, necessarily rare compared with the deterministic (and fast3) break-up of big liquid globules found at the clouds' base4 (hence their absence in the falling rain), as well as not always being efficient for break-up5: a collision may result in aggregation (forming a bigger drop), in nothing (drops just bouncing off each other), or in splash (forming fragments). Splash occurs under even rarer, very special conditions of trajectory alignment and relative velocities. Even at the scale of a burst drop, fragments are so dilute that they never interact. In any case, the effect of these decorative events is completely screened by spontaneous break-up, which in itself contains the whole spectrum of drop sizes, and coincides quantitatively with the Marshall–Palmer distribution, a widely accepted fit for the drop size distribution in rain6.

One swallow does not a summer make: collisions are one among the many phenomena in the rich zoology that accompanies rainfall7. Historically, collisions have been invoked for mediating aggregation, or break-up, or both8, but in spite of their fortuitous occurrence in ad hoc laboratory settings9, their rarity in nature disqualifies them as principal actors, whereas the spontaneous break-up model offers a quantitative understanding for the structure of rainfall3.