Summary: 

After swapping a few electronics that we broke while investigating the ALS glitches at EY, the glitches are still there, but not nearly as bad as they were.  We have had these glitches go away on their own before when nothing was done, so that might be what is happening.  Now they are at End X, and about as bad as they were at End Y for most of the week.  ALS still will not lock for longer than 10 minutes.  It might be that this problem will go away on its own, but will probably come back again.  

• After seeing that the behavior of the VCO at end Y tune mon changed around 6:30 UTC June 2nd, and based on Keita's alog last night, I swapped it for a spare I found in the EE shop this morning.  The unit I removed is S1200566, the unit I replaced it with is S12000562.  The VCO is no longer railing and the tune mon is near zero when the tune input is terminated.  The tune monitor BNC however has about 900 mV output even with the tune input terminated.  After this the arm would lock and I was not getting any VCO errors
• I found that the Seismic configuration was left in Large earthquake mode since Friday, and changed it to windy.  I found this because the arm transmitted powers were fluctuating more than normal for both arms, and also the oplevs showed extra noise since Friday evening.  Now the arms are both locked and quiet.
• At 1:08-1:23 local time, I started 15 minutes of locking the Y arm in green without automation (the PLL automation is on, not the PDH automation).  At 1:32-1:47 I did another 15 minute test with the only difference being that the PDH CM board was controlled by the tester, not the automation. At 1:50 I locked the arm using the automation, in the normal way. During all of these tests, there are a few glitches, but not as many and not as large as we were seeing mid week and last weekend. 
• I attempted to lock ALS, which didn't hold for more than a few minutes.  I then noticed that we are having the type of glitches we had been struggling with in the Y arm in the X arm now. (1st attached screenshot shows the Y arm and X arm both locked in the normal configuration with automation).
• I went to end X, and put the cable for the PDH signal into IN1 of the PLL board.  The X arm locked for green with in 1 polarity negative, and -4 dB of gain.  (For Y arm, we used positive polarity and 6 dB of gain).  No glitches in this configuration. 
• I also swapped the cables to substitute the spare FET demodulator for the PFD.  I added a 20dB attentuator to the RF cable into the demod, with this in I got a ugf of 27 kHz for an in1 gain of 13 dB. The autolocker works fine to lock this, you just have to force it to ignore the error message from the PFD by hitting the "Force" button. The arm was locked this way for 15 minutes starting at 0:03:30 UTC June 5th for about 10 minutes, the glitches were just as bad as the normal configuration. 
• I used Keita's set up from End Y to use the CM board tester at End X.  The cavity was locked this way from 0:47 UTC to 0:54 UTC June 5th, glitches are just as bad as the normal configuration.  

Things that are not the cause of glitches at EX:

From PLL bypass test:

• PLL servo board
• PLL servo board controls
• Laser
• PDH sensing, including diode, demod, and LO
• PLL automation

Other things that are not the problem:

• PFD (glitches still seen when it was swapped for a mixer)
• PDH servo board control cables swapped (so PDH control cables and automation are not the problem)
• loop instabilities (X end PLL UGF is 24 KHz with 47 degrees of phase margin, PDH Ugf is 5.6kHz with 56 degrees of phase margin).
• The beatnote is a clean sine wave, there are 8MHz sidebands on it but they are 40dB smaller than the carrier. 

Things that we haven't checked:

• PDH servo board, but this is not likely to be the problem since it wasn't at End Y and swapping it has some risks.
• VCO, we could try swapping this for an IFR but I wasn't able to find one in the shop today. 
• BBPD
• A problem on the light coming out of the fiber. it still seems like glitches on the PSL are unlikely because the glitches are not at the same time in the two end stations. 

I recommend that the first thing in the morning operators take the ISC lock guardian to locking arms green, wait there for 10-20 minutes, and look at second trends of the arm transmissions.  You can compare them to the screenshot I've attached here, if the glitches are gone or much smaller it would be worth doing an initial alignment and trying to lock to DC readout, if the glitches are still there it is not worth trying.