Where was the water buildup on Cassini's narrow-angle camera system? Did it have to remain heated continuously?

Question

Writing this answer led me to Wikipedia's Cassini; launch and cruise phase (1997-2003) which includes the following entries

0 May 2001 – During the coast phase between Jupiter and Saturn, it was noticed that "haze" became visible in the pictures taken by the narrow-angle camera of Cassini. This was first seen when a picture of the star Maia in the Pleiades was taken after a routine heating period.

23 July 2002 – In late January, a test was performed to remove the "haze" from the narrow-angle camera lens by heating it. Warming the camera to 4 degrees Celsius (39 degrees Fahrenheit) for eight days produced positive results. Later, the heating was extended to 60 days, and a picture of the star Spica showed an improvement of more than 90 percent compared to before the heating period. On 9 July, a picture showed that the removal procedure was completed successfully, which was announced on 23 July.¹⁶

¹⁶ NEWS - Press Release: Cassini Camera Haze is Removed July 23, 2002 https://web.archive.org/web/20061028020201/http://saturn.jpl.nasa.gov/news/press-releases-02/20020723-pr-a.cfm

which includes this image that I've zoomed:

Comparison images from Cassini camera

Now within two years of reaching Saturn, NASA's Cassini spacecraft took test images of a star last week that reveal successful results from an extended warming treatment to remove haze that collected on a camera lens last year.

The quality of the new images is virtually the same as star images taken before the haze appeared. In the most recent treatment, the camera had been warmed to 4 degrees Celsius (39 degrees Fahrenheit) for four weeks ending July 9. Four previous treatments at that temperature for varying lengths of time had already removed most of the haze. The camera usually operates at minus 90 C (minus 130 F), one of the temperatures at which test images were taken on July 9 of the star Spica.

The peak temperature reached during heating was +4 °C (normally -90 °C) so the haze was likely water, and the fact that the resulting image has a distinct annular appearance suggests it was due to somewhat monodisperse droplets on some optical surface.

Question: Where was the water buildup on Cassini's narrow-angle camera system? Does it have to remain heated all the time now? Did it have to remain heated continuously?

1. In which volume was the water trapped?
2. On which optical surfaces did it cause the problem?
3. Did the water get driven away or does the camera always have to run at +4 °C to run haze-free, or can it be operated at -90 °C?

Answer 1

I'll be referencing the 2006 paper Cassini Camera Contamination Anomaly:
Experiences and Lessons Learned by Vance R. Haemmerle and James H. Gerhard of JPL.

Contamination (not water) inside the CCD.

Contamination of optical instruments can occur at all stages of a mission, from instrument fabrication, transportation, test, integration with the spacecraft, launch and in-flight. Sources of contamination on the ground can range from a fingerprint to contact with ground equipment. In-flight contamination can result from loose items during launch, ejection of covers, outgassing of instrument and spacecraft materials such as water, organics, silicones and propellant and the in-situ environment such as atomic oxygen in low Earth orbit or interplanetary dust or ring material.

...

A review found two interesting and possibly relevant items. The first was that the radiators of the cameras had been resized with thermal blankets during integration with the spacecraft and had not been subsequently baked out. The thermal blankets used had been baked out, but of course the tape that was used was not. Secondly, a memo was found from 1995 from the contamination engineer to the thermal blanket engineer recommending that the CCD vent tube be extended. Due to a personnel change this action was not taken. This was a suspected contamination path.

...

The fact that a change occurred at −7°C confirmed to us that the contamination was indeed close to the CCD area (which was the only area to undergo a significant change in temperature), perhaps in the CCD package itself.

...

A meeting was held on November 16th to discuss the results [of C28]. It was decided to repeat the experiment at a slightly higher temperature. The only three choices to increase the temperature short of having both decontamination heaters on were to a) turn on the Replacement heaters, b) use the CCD performance heater fully rather than have it regulate the temperature or c) do both. The contamination experts wanted to reach the temperature at which the Stardust contamination was removed but that was not possible with the possible heater combinations. Water was ruled out as a contaminant since it should have evaporated. There was some concern that the contamination might be evaporating at the higher temperature and then re-condensing when the CCD cooled to −90°C again. It was decided to take images at a halfway point to check.

It was thought to be on the CCD window or filter.

Concerning the stellar images, the central peak of a star appeared normal and the intensity of the halo was only 1-2% the brightness of the central peak as seen in Fig. 6. However because of spatial extent of the halo, it actually contained a large fraction of the stellar flux (Fig. 7) - from 30% in the infrared to 70% in the blue and ultraviolet (Fig. 8). The size of the halo ranged from 5 pixels in radius in the ultraviolet to up to 20 pixels in the infrared (Fig. 8). Thus, this would have a large impact on the scientific return of the NAC. The properties of the point-spread function were consistent with the contamination by very small particles on a transmissive surface causing a diffraction pattern in images of point source objects. The surface involved could have been the filter assembly or the CCD window. An interesting side note was that Ellis Miner, the Cassini Science Advisor, related that his early graduate work involved measuring diffraction patterns in an entirely different field, medicine8.

The contamination appeared to be gone.

Up to this point, the NAC had used 22 of the budgeted 57 thermal cycles for the mission. The risk of using a thermal cycle must be balanced with the diminished returns of perhaps a reduction in PSF width of a few hundredths of a pixel that another decontamination might accomplish. It was decided that risk of further decontaminations outweighed the small possible future gains. A planned C34 decontamination was cancelled. The images were still taken since it was too late to change the sequence, but a real time command to turn on the heaters was not sent. Analysis of these images showed no change, as expected.

A new flight rule written by J. Gerhard, not to allow both Level 1 and Level 2 heaters ON at the same time (prohibiting going to +30°C again) was implemented. With the start of Saturn Tour, no further decontaminations are planned unless a problem reoccurs.