Where the tin-lead solder adheres to the copper of the lead (from plating) and to the pad (from hot air leveling or other pcb solder-coating method), an intermetallic of tin-copper forms. This is typically Cu6Sn5 or Cu5Sn6 and this intermetallic is what causes the solder to adhere to the copper pads and leads. During reflow, with the addition of tin-lead solders, the intermetallic grows. With other board metal finishes, other metallics form, such as AuSn when soldering components to boards with gold-flash finish (and tin-lead solder).
Over time, an intermetallic layer grows. The speed at which it grows is governed by the thermal excursions as well as the ambient temperature. One of the unfortunate properties of the intermetallic is that it is far more brittle than the tin-lead solder or the copper (or other metals involved). Over time, as this intermetallic layer grows thicker, it becomes the Achilles Heal of the solder joint in terms of integrity. Depending upon the stresses imposed on the assembly and the solder joints, this is what usually ultimately leads to joint cracking and failure.
Typically we measure intermetallic growth over months and years. However, during the reflow process, the intermetallic growth is somewhat accelerated due to the high thermal excursions (the highest the joint will ever see) and the fact that the tin-lead is in a molten state. With double-sided boards, there is additional intermetallic growth as the side A becomes liquidus again. If there is repair, the heat of the repair system can induce accelerated intermetallic growth in the joints being reworked and adjacent joints. (This is one of the key reasons we never perform touch-up for purely cosmetic reasons.)
While excessive intermetallic growth is difficult to measure and quantify on the shop floor, it is ultimately a cause of defects. Such growth can be as much as several microns during reflow. Therefore it is imperative that liquidus dwell time -the interval that the interconnection is above the melting temperature of the solder alloy, be kept as short as possible. It is also imperative that the board not be heated any higher than absolutely necessary and that peak temperatures are kept as close to T1 as possible.