Thank you for all the help, folks.
I soldered my second 4809 that I was trying in that clamshell onto a 48p breakout, and . . . it shows the same behavior as when it was in the clamshell. js will be happy that with the breakout my decoupling caps are a little closer at about 5mm and 9mm than they were before (as I said, I doubt that's the problem; only way to get them closer than with this dip breakout is buy a board or make one with kicad and a fab).
Also sadly, my review of that 5th photo schematic 4809 ultra-explorer thing doesn't reveal any problems with my setup.
Regarding "using Arduino as the programmer" (rather than STK600), it appears you are referring me to El Tangas' page about that. Maybe I could pursue that, though that's a bit unattractive. The only arduino I have lying around is a mega2560 one (though as was said, I could put the code into a breadboarded 328p or similar). It sounds to me like I need to recompile El Tangas' stuff if I want to pursue that on something other than 328p. Actually, I don't have the arduino 2560, must have loaned it to someone. It's either pursue El Tangas' way on big/fast enough AVR to host the code or wait for (or compile myself) avrdude improvements, though that would still be using the STK600 and might still fail mysteriously. Maybe somebody will figure out how to get STK600 to work. I would expect Atmel/MC would be interested in making it work, as they are selling the chip, and I doubt they want it to be this difficult/constrained as to how to program it and produce a running program. That or bag the 4809; there are plenty of AVRs that work great with STK600, though I was hoping to use the attractive patch bay on my clamshell board.