Dim before you go dark
A brownout borrows from electrical grids: instead of a full blackout, the system reduces nonessential function to save capacity. Under stress it temporarily turns off expensive features so the core keeps running.
This is graceful degradation made dynamic, controlled by an operator or by automatic load signals.
Feature flags as the control
Feature flags are runtime switches that turn behavior on or off without a deploy. They are the lever that makes brownouts and many resilience tactics possible.
- Disable a heavy recommendation panel when CPU is high.
- Reduce a page size limit to cut database load.
- Turn off a noisy background job during an incident.
Because flags change at runtime, you can react in seconds rather than waiting for a release.
Kill switches
A special flag is the kill switch: a single toggle that disables a risky or failing feature instantly. When a new feature causes an incident, flipping its kill switch is faster and safer than rolling back code.
Discipline required
Flags accumulate. Stale flags become hidden complexity and untested code paths, so teams must clean them up after they have served their purpose.
Key idea
Feature flags let you brown out expensive features in seconds, keeping the core alive without a deploy.