I've been able to get the SDRAM initialized correctly. (Once that was behind me, I was able to better diagnose and correct the "instabilities and fragile behavior" I was seeing.)
However, there are a few (I believe to be) problems with the board_memories.c source file (for both the Samsung and the Micron parts). When they do an EMSR1 (Extended Mode Register Set 1) cycle, the "acknowledge" it by writing to BA1 = BA0 = 1 (it SHOULD be BA1 = 0, BA0 = 1). These problems are in "step 18" of the Micron/Samsung initialization and "step 8" of the Samsung initialization.