Exposure time?

You are just hurting your image if you do drizzling so I'd quit that for starters. At B6 you sky brightness is the main limiting factor so anything above 300s is counterproductive, This said both the Iris and the Crescent are considered to be bright subjects but even then you would need many more hours to reach a reasonable SNR unless you keep your image scale down, like a LOT. 180s seems more like the thing all things considered. For not broadband subjects (such as the Crescent) get a dual NB filter such as the L-Extreme, ALP-T or equivalent from other manufactures to improve your signal gathering and reduce the sky brightness.

I'm not sure, read noise for the 2600 at a gain of 100 where HGC kicks in is a low 1.35 (e-RMS). This might have been good advice when shooting with CCD cameras but no longer holds with the new generation of CMOS cameras of which the 2600 is a good example. It's the light pollution that killing his S/N, forcing super long integration times to get anything. What's needed is narrowband filtering to block the light pollution. I shoot dual band OSC under bortel 8 skies. With that setup, 120 sec. subs work just fine on subjects like the crescent and that's at F/10.

I think it that it is fair to say that read noise in CCD's is generally greater than the current generation of CMOS sensors. I looked at some current CCD units in the 4 to 5 k range and the read noise ran between 9 and 11 e-RMS. This is a much higher value than the 2600. And yes, the read noise can kick down to a lower value if you set the gain properly using HCG mode. At a gain of 100 the read noise drops from an already low 3 e-RMS to about 1.35 e-RMS.  On my 585 sensor it drops from 3.75 e-RMS to 1.1 e-RMS. With read noise that low there just isn't any advantage to doing ultra long subs. You are only asking for lower resolution and more lost time from wasted frames.

Tony Gondola:
I think it that it is fair to say that read noise in CCD's is generally greater than the current generation of CMOS sensors. I looked at some current CCD units in the 4 to 5 k range and the read noise ran between 9 and 11 e-RMS. This is a much higher value than the 2600. And yes, the read noise can kick down to a lower value if you set the gain properly using HCG mode. At a gain of 100 the read noise drops from an already low 3 e-RMS to about 1.35 e-RMS.  On my 585 sensor it drops from 3.75 e-RMS to 1.1 e-RMS. With read noise that low there just isn't any advantage to doing ultra long subs. You are only asking for lower resolution and more lost time from wasted frames.

Its true that the value of the read noise was higher on older CCDs, but the problem still is still helped on CMOS by taking shorter exposures. And I see what you mean, I interpreted that as you saying that read noise only becomes a problem at certain exposure times, my bad. You're right in that there are certain gain modes that support higher dynamic ranges and less read noise, but I am assuming that is constant so we can discuss only the effect of longer exposures. Typically there are only one or two modes that most people will use anyway and they are local maxima of dynamic range. The contribution of the read noise on most exposures will be very small, nearly negligible if using reasonable exposure times, But why not use longer subs to reduce it?

If you are getting reduced resolution or having to throw away any significant number of frames by using longer subs, I would suggest that your guiding cannot handle it and that would violate the first criterion - I think that is covered by my original suggestion. There is no reason that resolution would be lost in a 10 minute exposure with good guiding vs. a 5 minute with good guiding as long as it's consistent. Unless youre lucky imaging, you will be sampling a very similar distribution of guide error which will dictate any mechanical loss to resolution.