fix(intercom): retry transient read failures

[altaywtf]

Summary

Harden live Intercom reads against transient provider/network failures seen during a 90d backfill attempt.

Changed

• increases the live CLI HTTP timeout to 90 seconds
• adds opt-in Intercom client retry attempts for request timeouts and HTTP 5xx responses
• keeps HTTP 429 handling in the syncer rate-limit layer only
• covers transient 502 recovery with a synthetic Intercom client test

Risks

• longer live waits when Intercom is slow or unhealthy
• retry behavior must not duplicate rate-limit retries

Verification

• go test ./internal/intercom ./internal/syncer ./internal/cli
• ./scripts/verify
• $HOME/.agents/skills/codex-review/scripts/codex-review --mode local clean after the 429 retry layering fix

Complexity

Reduced for operators: live backfills can survive transient provider failures without adding tenant-specific wrapper behavior.

---

Summary by cubic

Add opt-in retries for transient Intercom read failures and increase the live CLI HTTP timeout to 90s. This improves backfill resilience against timeouts and 5xx errors while keeping 429 handling unchanged.

• Bug Fixes
  • Added client retries for request timeouts and HTTP 5xx with MaxAttempts and RetryBackoff (CLI sets 3 attempts, 2s backoff).
  • Kept HTTP 429 handling in the syncer rate-limit layer (no duplicate retries).
  • Increased live CLI HTTP timeout from 30s to 90s.
  • Added a test that verifies recovery from a 502 response.

^{Written for commit 0931d71. Summary will update on new commits. Review in cubic}

[cubic-dev-ai]

No issues found across 3 files

_{Re-trigger cubic}

[Copilot]

Pull request overview

This PR hardens live Intercom “read” operations during backfills by adding opt-in retry behavior for transient failures and increasing the live CLI timeout to better tolerate slow/unhealthy provider responses.

Changes:

• Increased the live CLI HTTP client timeout from 30s to 90s.
• Added opt-in retry support in the Intercom client for request timeouts and HTTP 5xx responses (while leaving 429 handling to the syncer’s rate-limit retry layer).
• Added a synthetic test to verify recovery from a transient 502 during conversation search.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
internal/intercom/client.go	Adds configurable retry attempts + backoff for timeouts and 5xx, with proper body handling between attempts.
internal/intercom/client_test.go	Adds coverage ensuring a 502 on the first request is retried and succeeds on the next attempt.
internal/cli/cli.go	Increases live HTTP timeout and enables Intercom client retries for live sync runs.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.