Single events in radiation environment

In analog you mostly see transient perturbations. The amplitude,
duration, outcomes are highly variable with technology and topology.
Pulses can be stretched or squashed. "Analog" blocks can present
"digital" behaviors (like a PTAT locking up to a third high current
state if one particular transistor is kicked into saturation).

You start from "anything can happen" which includes all manner
of abnormal states, which you must root out and deal with to
make the assembly "self-righting" or at least surviving and awake
enough to follow a reset sequence from above.

Work backward from outcomes to causes and figure out what
can be "tolerated" with proper specs (& customer approval).
A test campaign on a prototype will probably reveal "stuff"
that any analysis failed to predict. Wheels within wheels, like.
I've seen multiple spins on complex parts, where peeling off
one layer only leads you to the next.