I’d actually say that (certainly for the blue curve) that the dark current is not linear in that plot – a proper analysis would show how close to a straight line it is. A few hundred ADU out may introduce a reasonable systematic error.
Anecdotally, using my QHY163M (same chip as the 1600) for deep sky imaging I’ve found that scaling (and even allowing Pixinsight to try “optimizing” darks) just doesn’t work properly. If you create a dark library for a few different exposures (10/20/30/45/60/90/120/180/300) then they can be reused for quite a while before needing rebuilding – just make sure the offset and gain are the same as for the lights.
A useful activity for a cloudy night or two…