Here are the screenshots, and as you can see the two sticks are identical. So it seems that 'reseating' the chips 'solved' the problem; perhaps they were loosely connected. But it's strange that MemTest86 did not detect any problem despite running it for nearly one day.
Well, I cannot seem to attach a screenshot to this post, but you'll have to believe me that the sticks were identical. The only field different in CPU-Z was "Channels #" which was "Dual" with both sticks in, and "Single" with each stick in separately. The DRAM Frequency was 535.8, the FSB
RAM was 3:8, the CAS# Latency (CL) was 7.0 clocks; the RAS# to CAS# Delay (tRCD) was 7 clocks, the RAS# Precharge (tRP) was 7 clocks, the Cycle Time (tRAS) was 20 clocks, the Bank Cycle Time (tRC) was 27 clocks, and the COmmand Rate (CR) was 1T. This was with the BIOS set to "AUTO" and Unganged (the default) for the Bios.
Advanced Chipset Setup:
Dram Frequency: AUTO
Dram Timing: AUTO
DCT Unganged Mode: AUTO
Share Memory Auto Detection: AUTO
I guess the lesson of this thread is that an improperly seated memory stick can pass all memory tests like the excellent Memtest86 test and yet still fail in practice--if indeed this problem was the result of an improperly seated memory stick.