The reference cavity was locked using a 21.5 MHz signal from a rack mounted board (FSS REF - Crystal Frequency Reference [http://www.ligo.caltech.edu/docs/D/D980353-C.pdf]). That turned out to have many problems and we removed it completely from the system circa 2008 with a fixed 21.5 MHz OCXO from Wenzel and an RF Amplifier (see [http://nodus.ligo.caltech.edu:8080/40m/986 elog]).
We got the design parameters for the reference cavity from [http://www.ligo.caltech.edu/docs/D/D980670-00.pdf D980670]
Some FEA of the LIGO/40m reference cavity using COMSOL: /FEA
See also Frequency Stabilization Servo.