Guiding glitches - again

I suddenly saw the same guiding glitches again that I had earlier this year. It results in images like this one:

What happens is that suddenly the scope moves into one direction by several arcseconds and then PHD2 slowly brings back the guide star.

I checked the usual suspects:

1. Cabling
   Checked if cables can get stuck somewhere. I even moved the cables out of the mount to check if they get stuck somewhere on the inside. No change.
2. RAPAS
   I always leave the RAPAS in (pure convenience). It could be that it pulls on the cable inside the mount. Removed it. No change.
3. Gear Meshing
   I wondered if I have too much backlash in my gears. I remeshed the gears. I remeshed them too tight (using a mallet and carefully bumping the gear box while tightening the screws). Guiding was pretty bad after that (it took PHD2 several pulses to get the guide star back - and then it immediately overshot. I also remeshed them too lose (resulting in more backlash) - which of course made guiding also bad. But: no change.
4. Periodic Error
   I normally don't use PEC as PHD2 can guide out the periodic error. But just to be on the safe side, I took another measure. Enabled PEC. No change.
5. Accessories
   I am normally hanging the keypad from the mount - when it's windy, it could bump into the mount. I also wondered if the Alnitak flip-flat box could be responsible for this. Removed both of them. I checked the counterweight bar and the counterweights. I checked the retractable dew shield of the TOA-130 and tightened its screws. No change.
6. Screws
   Checked all the screws an tightened them (scope rings, dovetail, saddle plate, mount...) No change.
7. PHD2 error
   I also checked if maybe PHD2 has an error and drives the guide star out. But as it is clear from this image, PHD2 doesn't make corrections until after the glitch occurs:

Next, I wanted to check if there is something in common between these glitches (happen at the same time, in the same location...) I went through my PHD2 log files and searched where I had these in the past. They are pretty easy to spot. A normal guiding graph looks like this:

You can see the regular dithering events where the guide star wanders out by up to 4.5 arcseconds. And also times when guiding stops and then resumes - either focusing or a meridian flip.

When these glitches happen, it looks like this:

This event is much larger then the 4.5 arcsec dithering event. And it occurs in the middle of imaging.

I created a table with all these glitches from the past months:

	RA	DEC	Date	Time	Altitude	Azimuth	Direction
M101	14:03	54:16	4/22/2014	23:35	66:58:00	35:20:00	RA
M101	14:03	54:16	4/23/2014	23:25	66:15:00	36:40:00	DEC
M101	14:03	54:16	4/27/2014	23:21	67:36:00	33:58:00	DEC
M101	14:03	54:16	4/28/2014	23:39	69:51:00	27:39:00	DEC
M101	14:03	54:16	4/29/2014	23:48	70:57:00	23:04	DEC
Crescent Nebula	20:12	38:23:00	5/1/2014	2:24	40:08:00	67:45:00	RA
Crescent Nebula	20:12	38:23:00	5/2/2014	3:26	73:07:00	52:20:00	RA
Crescent Nebula	20:12	38:23:00	5/5/2014	1:44	50:07:00	72:13:00	RA
Crescent Nebula	20:12	38:23:00	5/8/2014	1:39	36:02:00	65:43:00	RA
Crescent Nebula	20:12	38:23:00	5/10/2014	1:12	33:17:00	64:17:00	RA
Tulip Nebula	299:51:00	35:28:00	8/8/2014	22:40:32	42:14:00	35:13:00	DEC
Tulip Nebula	299:51:00	35:28:00	8/11/2014	2:59:51	64:32:00	356:13:00
Tulip Nebula	299:51:00	35:28:00	8/12/2014	2:29:00	63:50:00	10:57	DEC
Tulip Nebula	299:51:00	35:28:00	8/12/2014	3:07:00	64:37:00	358:43:00	RA
Tulip Nebula	299:51:00	35:28:00	8/13/2014	21:30:00	37:49:00	34:38:00	RA
Tulip Nebula	299:51:00	35:28:00	8/14/2014	2:14:00	64:13:00	7:50	RA

There are a lot of failures at altitude 65-70 degrees. Both in RA and DEC direction.

Also, the above images were taken with various configurations:

• the OAG and the new guidescope
• both cameras: the H694 camera (and SX filter wheel) and the new ML 16070 camera (with the FLI filter wheel)
• with the clamshell holder and the new scope rings
• the Robofocus focuser and the FLI Atlas focuser
• the Flattener and the Reducer

So, it can't really be anything of those.

What's left:

• The Rotator (connection to the imaging train and connection to the stock focuser)
• The Takahashi focuser (though with the FLI focuser I clamped it down)
• The connection of the Takahashi focuser to the optical tube
• Something inside the OTA (lenses or such)
• The mount itself
  The position in which the glitch mostly occurs, is between these two positions:
     =>   

It looks as if the center of gravity shifts here from the left to the right side. I.e. if there is anything loose on the mount it would sift slightly from left to right. That could also explain why I see DEC or RA errors - if it would be the RA (or DEC) axis itself, then the error would be primarily in that axis.

I checked and rechecked all screws and connections (legs, mount plate, mount screws...) but couldn't find anything yet.