Regarding error bars, I don't know of any examples treating objective metrics but there's some discussion of confidence intervals for subjective viewing results in section 4 of doi.org/10.1186/s136... which might be useful. Mathias Wien might have written more about this elsewhere also.
Big fan of shrinkray! I’ve been using it to shrink binary testcases
www.frankplowman.com/posts/video-...
I think might be more of a display/colour space problem than a video coding problem. Regarding the coding aspect though, VVC's LMCS feature tries to address this by allowing the encoder to allocate more bits to a particular part of the luminance curve sigport.org/sites/defaul....
+1
Imperative ISOBMFF parsing code is very verbose, and the format is consistent enough at the high level that it feels as if some declarative approach would be better. In practice, I’ve found there are too many inconsistencies at the lower levels to design useful abstractions over.