The thing about images is that compared to text, they just suck to design with.
A few basic assertions:
- Images should be large enough to provide value, otherwise its just wasted space.
- The content of the images determines the minimum size it must be to add value.
So IF we are in a context where images must be "large" in-order for them to add value, then seeing more than one image will mean the user must be inconvenienced somehow, as the screen space / real estate will not be enough to view more than one image at a time.
- User will have to click to see next
- User will have to wait to see next
- User will have to scroll to see next
So now that the basic heuristics are out of the way. Because this issue has come up for you, its safe to assume that these images need to take up a significant portion of layout. So the question is, what inconvenience do you feel is the least abhorrent.
It sounds like scrolling is the preferred inconvenience, but you also value the fixed height of the carousel. Luckily you don't have to choose (depending on your layout).

download bmml source – Wireframes created with Balsamiq Mockups
You can have a fixed region that also scrolls. Potentially you could also have it 'auto-scroll' to the next image with a timer, which effectively makes it a carousel.
There are a zillion ways to do a carousel, and every interaction should be determined based its value.
Thumbnails also have value as they give a preview for non-linear navigation, but its about your use case. And you could so have this with scrolling using waypoints. Its all possible.
Really my point is that carousel vs scrolling is a false choice, and you should be able to do better than sticking within one of those two rigid paradigms.
Agreed with @Paul that NN has great guidelines for the usability of carousels.
You can do all the A/B testing you want, but first it would be wise to ensure you aren't testing a false binary.
"It sends a premium feel, like the product has been worked with care and it's not a gimmick, keep in mind this are rather expensive products."
This is also an important point. If your design is beautiful and the animation feels slick, regardless of what route you choose; its going to benefit from aesthetic usability principle.
Because of this, I feel that the visual treatment is going to be equally if not more valuable for your use-case, than the layout of interaction scheme.