/ripple

Cascade Analysis

Purpose: Detect second-order and emergent effects beyond direct dependency chains. Read when: Cascade analysis is triggered (≥3 service boundaries, bidirectional deps, shared resources ≥3 components, or risk ≥7).

Trigger Conditions

Effect Types and Detection

1. Feedback Loops

A affects B, B's response amplifies or dampens A.

Detection:

1. Build bidirectional dependency graph from VERTICAL analysis
2. Identify cycles: A → B → A (direct) or A → B → C → A (transitive)
3. Classify: amplifying (positive feedback) or dampening (negative feedback)
4. Assess stability: amplifying loops are higher risk

Examples:

• Cache invalidation triggers reload, reload triggers cache invalidation
• Retry storm: failure → retry → overload → more failures
• Auto-scaling: load increase → scale up → connection pool exhaustion → more load

2. Cascading Failures

Sequential failure propagation across boundaries.

Detection:

1. From VERTICAL L2+ results, identify cross-boundary dependencies
2. For each boundary crossing, assess: what happens if this dependency fails?
3. Map failure propagation paths: single failure → downstream failures
4. Check for circuit breakers, bulkheads, or isolation mechanisms

Failure propagation patterns:

• Synchronous chain: A calls B calls C — C failure blocks B blocks A
• Resource exhaustion: A failure causes B to retry, exhausting B's thread pool
• Data corruption: Invalid data from A propagates through B and C before detection
• Configuration cascade: Config change in A invalidates assumptions in B, C

3. Emergent Behavior

Combined changes produce unexpected system-level properties.

Detection:

1. From HORIZONTAL analysis, identify pattern interactions
2. Check: do two independently-valid changes conflict when combined?
3. Assess behavioral invariants that span multiple components
4. Look for implicit contracts (timing assumptions, ordering guarantees)

Common patterns:

• Two performance optimizations that individually help but together cause thundering herd
• Independent schema migrations that create inconsistent intermediate state
• Feature flags that interact unexpectedly when both enabled

4. Resource Contention

Multiple affected components compete for shared resources.

Detection:

1. Map shared resources: databases, caches, queues, connection pools, file locks
2. For each shared resource, count affected components from VERTICAL analysis
3. Assess current utilization (if available) and projected change
4. Flag resources where affected_components ≥ 3

Shared resource categories:

• Database connections / connection pools
• Cache (Redis, Memcached) — key space collisions, eviction pressure
• Message queues — consumer lag, partition rebalancing
• File system — lock contention, disk I/O
• External API rate limits — shared quota consumption

5. Temporal Cascades

Effects that manifest only under specific timing or ordering.

Detection:

1. Identify async dependencies in affected scope (event handlers, queues, cron jobs)
2. Check for ordering assumptions: "A always completes before B starts"
3. Assess migration/deployment ordering dependencies
4. Look for race windows created by the change

Risk signals:

• Eventually-consistent data read by strongly-consistent consumers
• Deployment ordering requirements between services
• Cron job timing assumptions invalidated by the change
• Event ordering guarantees broken by async refactoring

Integration with L0-L3 Analysis

Cascade analysis extends each depth level with second-order effect detection, but does NOT replace the standard confidence degradation (L2 = medium, L3+ = lower confidence).

Output Format: Cascade Risk Map

Append to the standard impact report:

## Cascade Risk Map

### Feedback Loops Detected
| Loop | Type | Stability | Risk |
|------|------|-----------|------|
| [A → B → A] | Amplifying | Unstable | HIGH |

### Cascading Failure Paths
| Origin | Path | Boundary Crossings | Mitigation |
|--------|------|--------------------|------------|
| [Component] | [A → B → C] | [count] | [circuit breaker / bulkhead / none] |

### Emergent Behavior Risks
| Interaction | Components | Risk | Detection Difficulty |
|-------------|------------|------|---------------------|
| [description] | [A, B] | [HIGH/MED/LOW] | [HIGH/MED/LOW] |

### Resource Contention Points
| Resource | Affected Components | Current Utilization | Projected Risk |
|----------|--------------------|--------------------|---------------|
| [name] | [count] | [%] | [HIGH/MED/LOW] |

### Temporal Cascade Risks
| Assumption | Invalidated By | Impact | Mitigation |
|------------|---------------|--------|------------|
| [ordering assumption] | [change] | [description] | [recommendation] |

Routing After Cascade Analysis