The issue seems to be that if you write to MISO, MOSI or SCLK (P1.5 - P1.7) within 500ms of the device coming out of reset the device becomes unprogrammable via the SPI port. The errata sheet on Atmel's site states this is because the pins are supposed to tri-state waiting for the programmer to do something. I've found this to be partially true, ie programming fails but if you try, and try, and try again it will succeed.

Advice I've read elsewhere suggests a delay of 750ms before writing to these pins. I wait one second...

I don't think flash corruption is related to SPI, only to 'brownout' conditions. I don't know if your supervisor chip is suitable - perhaps one of the many hardware experts on this forum would comment?