As I feared, I am just jinxed when it comes to the hand-soldering phase. I really wish I knew where the problem was. I used the same technique as the practice board: lay down some solder on the pad without flux, put the pins in, then flux and solder.
I ranged from a couple of really weak joints, and with most of them the pad just disintegrated as if it was being burnt. Once that happens, solder won't adhere no matter what you do. If you look closely, you can see some of this happening on the reflowed components (top pad on the resistor, pads closest to the crystal on the caps). Those joints seems to work in spite of the problem, but the bar is lower vs a through-hole part.
I'm pretty sure my iron isn't too hot, as once it was clear the board was toast, I switched back to the original solder as an experiment. The iron could barely make that solder wire melt, whereas the new wire melted more readily. If it were running too hot you'd think that wouldn't be the dynamic.
So, now I have to start again. (heavy sigh)
The only difference I can think of is the practice board has really large oval pads. This gives the heat someplace to go, reducing the risk I'm burning out the ink. It also means if the solder start to leach the silver out of the ink, there's still pad left for the solder to adhere. I'm also being extremely stingy with the flow rate in order to print at this level of fine detail, which causes the height of the trace (and hence the volume of ink) to be less than normal.
Maybe it's time to print a sample board with ring pads and oval pads and see what happens...