Summary / What changed

• Harden transient network detection in src/infra/unhandled-rejections.ts.
• Traverse wrapped error shapes (cause, original, error, data, errors[]) before deciding fatal/non-fatal.
• Detect transient network codes from nested fields and message text (for wrapped Slack request errors).
• Add regression coverage in:
  • src/infra/unhandled-rejections.test.ts
  • src/infra/unhandled-rejections.fatal-detection.test.ts

Why

A short DNS failure (EAI_AGAIN) should not terminate the whole gateway process.
Before this change, wrapped request errors could bypass transient detection and trigger process exit, causing restart loops and channel downtime.

Tests

• pnpm exec vitest run src/infra/unhandled-rejections.test.ts src/infra/unhandled-rejections.fatal-detection.test.ts --config vitest.unit.config.ts

Real-world validation

• Reproduced and validated on a live Linux VPS on February 23, 2026.
• During transient DNS instability, gateway no longer exits on wrapped request errors after this patch.

AI-assisted disclosure

• AI-assisted: yes (Codex).
• Testing degree: targeted unit/regression tests + live validation.

Issue

Fixes #24581

Greptile Summary

Enhanced transient network error detection to prevent gateway exits on wrapped DNS/network failures by implementing breadth-first traversal of error wrapper structures (cause, original, error, data, errors[]).

• Traverses nested error shapes to find transient network codes at any depth
• Adds message-based detection via regex patterns and text snippets for wrapped errors where codes are buried in messages
• Specifically handles Slack WebAPI request errors that wrap EAI_AGAIN and other network codes in .original field
• Maintains proper handling order: AbortError → Fatal → Config → Transient → Generic
• Comprehensive test coverage validates both new wrapper detection and existing behavior

Confidence Score: 5/5

• Safe to merge with minimal risk
• Implementation is well-structured with proper circular reference prevention, multiple detection layers for robustness, comprehensive test coverage including edge cases, and maintains backward compatibility while solving a critical production issue
• No files require special attention

_{Last reviewed commit: de2d637}