Batch Optimization Failures
Frequency:Common
Category:Operations
Processing Items Individually When Batching Would Be More Efficient
Cost Efficiency
15 patterns for this issue
Caching Failures
Frequency:Common
Category:Operations
Redundant LLM Calls for Identical or Similar Requests
Context Stuffing
Frequency:Common
Category:Operations
Overloading Context with Irrelevant Information
Cost-Quality Tradeoff
Frequency:Common
Category:Operations
Cost Optimization Degrades Agent Quality
Expensive Model Cascade
Frequency:Common
Category:Operations
Agent Escalates to More Expensive Models for Tasks That Don't Require Them
Infinite Loops
Frequency:Occasional
Category:Operations
Agent Gets Stuck in Infinite Retry Loops
Model Selection Waste
Frequency:Very Common
Category:Operations
Using Expensive Models When Cheaper Alternatives Suffice
Resource Exhaustion
Frequency:Common
Category:Operations
Agent Exceeds Allocated Resources Before Completion
Retry Amplification of Costs
Frequency:Common
Category:Operations
Exponential Retry Logic Multiplies Costs Beyond Original Request
Retry Storms
Frequency:Common
Category:Operations
Aggressive Retries Multiply Costs
Step Repetition
Frequency:Common
Category:Operations
Agent Repeats Same Steps Without Progress
Token Budget Overflow Attack
Frequency:Occasional
Category:Operations
Attacker Intentionally Triggers High-Token Operations to Exhaust Cost Budget
Token Explosion
Frequency:Common
Category:Operations
Exponential Token Usage Growth
Unnecessary Tool Calls
Frequency:Common
Category:Operations
Agent Calls Tools When Not Needed
Verbose Reasoning
Frequency:Very Common
Category:Operations
Excessive Chain-of-Thought Output
Token and API cost management
Failure Patterns
Total: 12 patterns