[jase (Jason)]

Mark is correct, but it should be made clear as to why you'd need to rotate the camera post a meridian flip as the comment that your subs end up upside down is irrelevant. In fact, I'd go as far as saying that subs taken on the east, then the west side of a mount can actually benefit the imager as it will place chip anomalies (dead pixels, column defects) on different parts of the image making them easier to remove through data rejection algorithms.

The primary reason why you rotate the imaging train post flip is to pick up the same guide star when guiding on axis. Rather important when you're operating at long focal lengths. For the regular imager, particularly one using a wide field instrument this is a mute point as guide stars are typically plentiful for on axis guiding. So in this case, you wouldn't bother rotating your imaging train.

Automated data acquisition tools such as ACP, CCDAutoPilot etc will take this into account;
"With rotator support, if the telescope is a GEM, half of the flats in each set are taken with the rotator at the specified PA and the other half are taken 180 degrees from the specified PA. This allows making flats for the common case where the target crosses the meridian, invokes ACP's auto-flip feature, and thus has a 180 degree rotation in the middle of acquiring its images. The west and east images require flats with rotations 180 degrees apart."

I have heard of people rotating the flat manually to address this. We should note that we're talking quite an advanced technique here and it can generate considerable workload...
"Note that use of a rotator for multi-target imaging can quickly lead to needing far more flats than you can produce. This is the main disadvantage of using an internal guider and rotator. Careful planning of multi-target runs can help. For example, if you are going after three targets over three nights, you can take one set of flats for each target on each of the three nights at dusk and dawn. Take all of the flats for a particular sky position angle (for one target) in one dusk or dawn run to eliminate rotator slew time detracting from the time available for flat acquisition."

No one said producing flats was straight forward. Once you've done it a few times, familiarity kicks in and it becomes routine. Start off with lightbox or tshirt flats. Skyflats need further consideration as the sky is dynamic in nature. At dusk its constantly getting darker, at dawn, brighter. This needs to be factored in if you're trying to hit the target ADU count for each flat sub illumination (20%-30% well depth).

Flat Darks Addition: Don't forget those with cooled CCD cameras that can take bias frames have the ability to scale a dark frame to match a light frame (or in this instance a flat frame). In essence, subtracting a bias frame from a dark frame provides a thermal frame to match the light frame exposure. The act of scaling darks can work exceptionally well if you've got flat frames with varying exposure times. For example a narrowband Ha filter does not pass as much energy as a green filter, hence would require greater exposure time to reach the target ADU. Dark scaling would allow you to take a set duration dark say 300s and scale it to match the light frame providing its under the 300s duration.

Cheers