I found this thread based on a thread I had posted on the Spark Fun Electronics forum here:http://tinyurl.com/yxzfyl
I was having trouble connecting an accelerometer break-out board for the LIS3LV02DQ over SPI. By itself, it worked fine. But, when I put a second device on the SPI in parallel, the LIS3LV02DQ started sending spurious, but consistent, data. (By that I mean that the data was wrong, probably off a bit or two or three, but consistently so. The "who am i" register is supposed to send 0x3A and would only send 0x30 or 0x38 when SPI for the device went wrong.)
If I connect the LIS3LV02DQ "closest" in parallel to the Arduino pin 13 (the SPI Clock), it works perfectly. With a second device (a DS1306 RTC) also in parallel, the LIS3LV02DQ only worked if I had a jumper from Arduino pin 13 to the LIS3LV02DQ clock in, _then_ to the DS1306 clock in. The DS1306 worked in either arrangement.
Based on this thread, I would normally suspect this issue with pin 13, but I have an older Arduino without the LED there. But, I'm still suspicious of an electrical design issue there, perhaps in combination with the LIS3LV02DQ being especially sensitive to a weak clock. (BTW the LIS3LV02DQ is a 2.6-3.6v device — perhaps a weak clock, sunk even further with high impedence inputs nearby, tosses the device into the weeds.)