@tnt: I just did a test about ten minutes ago where I changed the

sections.lds

for the linker to move the subroutine call stack from SRAM to DFFRAM, and suddenly I was able to run a program with the chip powered at 1.8V and everything in the CPU (not the GPIOs) worked fine. So it's the SRAM that fails to operate at 1.8V and has to be run at 1.6V. My test does not indicate how the SRAM fails or if it does so on a read or a write.