Its complicated. It certainly is a function of how long the short occurs as
to burnout, and many device outputs are not capable of developing enough
current in short to cause thermal related damage. But a short can cause
internal power rails to collapse and cause logic errors in onboard logic
state machines that may or may not recover when power is restored
if power up ramp rate not met, usually in datasheet. If outputs buffered
onboard with high drive outputs you can get bond wire failure and or
power rail failure. Even local hotspot substrate damage. Vendors have,
rel departments, over the years documented a lot of this.
NASA uses in space applications EEPROM, but seems to have developed
their own R/W controllers to bolster reliability. Lots of papers in NASA
archives on non vol memory technologies and their reliability.
My experience, todays devices, is shorts outside the power rails most often
take down parts. The short to Vcc when output is low less so, and the short to
ground when output driving high rare for permanent damage.
It very much depends on device sizing on the outputs. There has been an
evolution where once high source drive current << low sink current is now symmetrical.
So my knowledge is a tad dated, Contact rel departments, often they can be of great
help.
Regards, Dana.