If by "gull wing" you are referring to this packagehttp://workmanship.nasa.gov/lib/insp/2%20books/links/sections/707%20G...
then you might get some hints from NASA Workmanship page 8.02 under
Conformal Coating and Staking (Bonding) from herehttp://workmanship.nasa.gov/lib/insp/2%20books/frameset.html
My guess is that under vibration the body of the package is moving out
of sync with the pcb vibrations - and with differing amplitudes -
therefore stressing the leads and the soldered joints. Bonding the
body of the IC to the board in accordance with NASA recommendations
may help in ensuring the package moves in sync and with the same
amplitude as the pcb vibrations. A thicker pcb should also help
suppress uneven modal vibrations in the pcb itself.
The problem is more likely to be the flex of the PCB than anything
else. The resonant frequency of the package vs PCB mas spring system
will be very high. You have to get near the resonant frequency before
you get large deflections.
A PCB supported by standoffs only at the corners can get into many
very bad resonance modes. There is the obvious simple one. The PCB
can form an "S" shape parallel to either edge or running from corner
to corner in either direction. The standoffs can lean back and forth
allowing for modes where the PCB shakes right left, front back and
twistwise.