All tests OK including the Rev.A and Rev.B checks, though, neither of the last two are honored in the release version. It's kind of interesting though, I never knew the white line was not supposed to be there. Does this apply to the US version as well?
FWIW, I did a quick test and traced some of the mid-screen changes it makes after the ISR call.
From earliest to latest PPU cycle:
73115-73117 $E001 W
73148-73156 $2002 R
73175-73192 $2006 W dual writes for setting up $0B00
73198-73204 $2007 R from $0B01 or $0B02 depending on scroll-address clocking interference, $2001.18 flags are both set!
73226-73234 $2000 W has no effect, ctrl-reg #0 == data == $A8
73238-73246 $2002 R
73261-73282 $2005 W dual writes for setting up $73A0
73286-73292 $E000 W
73319-73327 $8000 W has no effect, ctrl-reg #0 == data == $47
More updates if/when I find out more...
It's possible that the glitch at the top and bottom parts of the screen in CV3 are there on the real thing as well, probably not as visible when viewing it on an NTSC television though. It seems to have been partially corrected in the PAL version.
Btw, nice work Belmont!