Resilience is not retry logic.

Basic retry:

for attempt in range(3):
    try:
        call_service()
        break
    except TimeoutError:
        sleep(backoff(attempt))

for attempt in range(3):
    try:
        call_service()
        break
    except TimeoutError:
        sleep(backoff(attempt))

for attempt in range(3):
    try:
        call_service()
        break
    except TimeoutError:
        sleep(backoff(attempt))

Retries mask symptoms. They do not address systemic instability.

Designing for Failure Domains

Segment infrastructure into isolated domains:

services:
  - api
  - billing
  - auth
failure_domains:
  - compute
  - storage
  - network

services:
  - api
  - billing
  - auth
failure_domains:
  - compute
  - storage
  - network

services:
  - api
  - billing
  - auth
failure_domains:
  - compute
  - storage
  - network

Isolation prevents cascade amplification.

Adaptive Recovery

Recovery should evaluate state:

if state.dependency_health["db"] == "down":
    route_to_replica()

if state.dependency_health["db"] == "down":
    route_to_replica()

if state.dependency_health["db"] == "down":
    route_to_replica()

Resilience is proactive adaptation, not reactive recovery.