The diffracted power for the ISS was going from 0 to 20% on the MEDM screen and both Travis and
Jeff were having problems getting the servo to lock.  I tried the following steps to see what the
behaviour was:
 - the autolock was dis-engaged
 - decreased the gain from 16.0 dB to 3.0 dB -> no effect
 - increased the magnitude of REFSIGNAL from -2.02 to about -2.30 -> no real improvement
 - turned off the integrator -> no real improvement
 - turned off the second loop -> no improvement

Then it was onward to some signal tracing ... The ISS acousto-optic modulator (AOM) gets its 80 MHz
RF input from one of the oscillators in the ISC.  According to the ISC_CUST_REMOTERACKS MEDM screen
it'd be the VCO in slots 24 and 25.  Its output is amplified by the amplifier in slot 23.  The output
of the RF amplifier was relatively stable, so it was not the cause of the behaviour observed with the
diffracted power.

From the ISS MEDM screen, there were no indications of any signal injections.  Switched AUTOLOCK off
and went to manual mode.  Turned off the integrator.

If it were a loop oscillation, reducing the gain slider ought to help.  The gain slider was reduced
back down to 3.0 dB.  The output AC of PD B indicated that the loop was sporadically locked but the
diffracted power was still varying wildly.  Moved REFSIGNAL in the other direction, dropping it as
low as ~ -0.8.  It was here that the loop showed signs of it being solidly locked according to the
output AC of PDB.  However the diffracted power was around 50%.  REFSIGNAL was then slowly adjusted
to bring the diffracted power to about 16%.  The servo gain was then increased back to its nominal
value of 16 dB.  Everything seemed fine at this point, so the reference signal was adjusted to -1.98
which brought the diffracted power to ~ 7%.

The integrator was engaged, which resulted in a decrease in the diffracted power.  The second loop
was turned on, but not engaged, without any detrimental effects.  AUTOLOCK was restored.
(see ISS-Summary1.png)

At one point the diffracted power did vary and this coincided with the IMC dropping out of lock
(see ISS-medm1.png and ISS-Summary2.png).  One can see in the summary plot three small peaks
in the diffracted power, AOM driver monitor output and PD B output that coincide with when the IMC
dropped out of lock.

So now the question remains as to why this was happening.  The location of the photodetectors for the
first loop ISS is after the pre-modecleaner.  Obviously in order for them to see light, the pre-modecleaner
must be locked.  Which in turn relies on laser frequency excursions to be slower than the response
time of the PZT and its high voltage driver.  Left to itself, the laser frequency does not change
much however it is driven by the need for the IMC to be resonant.  This is split between the frequency
shifting AOM in the FSS and the IMC suspensions.  So if the IMC length changes, for whatever reason,
one would expect to see changes in the ISS.  A quick inspection of H1:IMC-L_OUTPUT to me suggests that
it spent a fair amount of time swinging (see IMC-L.png).