> Can we somehow effectively, assuredly, in a way that citizens can trust, limit the surveillance activities to legitimate national security and lifesaving actions?
Unlikely, every system of checks and balances requires someone to set up and enforce those checks and balances so that person can then circumvent them.
It's the classic "How do you protect yourself from a rogue admin"
There are steps you can take (use a quorum etc) but against a malicious actor with sufficient power there isn't much of a defence.