Cut & paste from en email I just sent:
[I] predict that OTA flashing turns out to be an excellent test case for the problem described here in the forums:
I have not yet dug into exactly what mutex is required around the SPI CS assertion, but there is enough smoke surrounding this issue to expect fire in the same area.
I actually don’t understand what the CC3000 driver would be doing in an irq context, since the driver seems fully polled; but if the flash driver uses interrupts, that will be enough to cause problems.
I fully expect that if you were to hang an SPI analyser off the bus when this problem happens you will see mangled SPI transactions that are resulting in the send() return value of zero.