A small leakage current path can exist between the VCCINT and VCCAUX voltage rails in Spartan-6 devices during a power-down / power-off cycle. This situation only exists on a very small subset of devices. It can be triggered when VCCAUX is powered-off prior to VCCINT, and the VCCAUX voltage lowers to approximately 0.75V. The leakage can be as high as 17mA, but is much more typically less than 3mA or completely non-existent.
If there are no other active or passive devices on the board sharing the VCCAUX rail, or if the current loading of other devices is too small, the VCCAUX rail can actually "hang" around the 0.75V level due to the current being fed from the VCCINT voltage rail, until the time that the VCCINT power rail is actually powered-off. If the VCCAUX voltage is used to directly control the enable pin of the VCCINT rail voltage regulator, this can cause the hazardous situation where the VCCINT regulator never shuts down.
The Spartan-6 device will naturally load the VCCAUX voltage rail, drawing current and lowering the voltage to at least 0.9V before the potential VCCINT current leakage path can be triggered. So, the suggested work-around if the above mentioned hazard situation is implemented on a board is to change the enable circuit of the VCCINT regulator so that enable pin of the VCCINT regulator will trigger to shut off its output when a VCCAUX voltage level of 1.0V (or higher) is reached. Typically, this can be accomplished by using a resistor divider network on the Enable pin.
Alternatively, a resistive, deterministic load can be placed on the VCCAUX rail that is either always present, or enabled only during a power-off event. This load would need to "bleed" off the VCCINT current being leaked to the VCCAUX voltage rails, and lower the VCCAUX voltage low enough to trigger the VCCINT regulator's enable.