Wikipedia is cute and all, but I think I'll reference something with a little more qualification..
NASA Jet Propulsion Laboratories, for instance..
http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/20708/1/98-1767.pdf
These are the conclusions after testing the way that solder joints fail on a variety of different component packages and a variety of different PCB finishes.
-------------------------------
Conclusions
For plastic packages, crack initiation, propagation, and failure occurred at either the package or
board interfaces for sections with or without voids. This was true for A or B cycle conditions.
Generally, voids were concentrated near the package interfaces. There appeared to be no crack
propagation among the voids, except for the voids interconnected at the interface.
For a ceramic assembly failure (See Comment Below), brittle failure was observed for Nickel-Gold surface finishes. OSP and
HASL showed ductile failure through eutectic solder joints. For plastic packages, there was no
distinction between the three surfaces
-------------------
The ceramic packages tested were very odd - from the document:
Ceramic packages with 625 I/Os and 361 I/Os were also included in our evaluation. Ceramic
packages had high melting solder balls (90Pb/lOSn) with 0.035 inch diameters. These balls were
attached to the 'ceramic substrate with eutectic solder (63Sd37Pb). At reflow, package side eutectic
solder and the PWB side eutectic paste was reflowed to provide the electro-mechanical interconnects.
---------------------------------
The high melting point solder is nothing like standard leaded or lead-free. It's VERY heavy in gold content. Nothing like that is used on the 3 XM-L Board
PPtk