Bernard,
Before I had my current set-up, I used Lenny Eiger (Petaluma, CA) when I needed the best possible drum scans of analog film. IIRC, Lenny felt that 8000 ppi was effectively the limit to extracting useful information from even the finest grain films, and interestingly enough, that is also the maximum resolution at which my Imacon 848 scans 35mm film. With my current Arca-Swiss/Rodenstock HR Digaron/Trichromatic set-up (christened, tongue-in-cheek, "Lux ex machina"), very precise film positioning allows me to use 11,530 pixels on the long dimension of the sensor for the image of an unmounted 35mm transparency. Since the Trichromatic sensor is 11,608 pixels wide, that's actually 99.3% of the sensor width, and an effective resolution of 8,138 pixels per inch.
I could, potentially, install the 50cm bellows in the f-metric, nearly double the magnification and digitize a 35mm transparency in two stitched sections, but so far I haven't felt the need, and I'm frankly doubtful that it would improve my results. I'm effectively imaging dye clouds, and I can already see every individual particle (at least I think I can!). Whether you capture a single frame or stitch, establishing nearly perfect parallelism is a critical and exacting task, and I will note that I use a Versalab Parallel laser and a 4-6λ first-surface mirror from Edmund Optics to do that job. 8x10 film at 8000 ppi, now that would be some file!