Ensure JSON logging fallback never drops records

[groupthinking]

Harden the _format_json fallback so serialization failures, forged class values, oversized ints, and non-finite floats still emit a valid JSON record.

[linear-code]

GRV-407

[github-actions]

👋 Thanks for opening this issue! It has been automatically labeled for triage.

[groupthinking]

Duplicate of #1525 ("residual JSON fallback can still drop records after #1491"), which covers the same four gaps — serialization failures, forged __class__ values, oversized ints, and non-finite floats defeating the _format_json fallback. #1525 was already fixed and merged via #1515, closed moments before this issue was opened. (See also #1501, still open, tracking the oversized-int/raising-__str__ holes specifically, which #1515 also addresses.) Closing as a duplicate; please reopen with a repro against current main if a gap remains.

---

Generated by Claude Code

[coderabbitai]

⚠️ Possible Duplicate Issue(s)

• GRV-394
• _format_json still loses records after #1491 — four holes, nine PRs, none merged #1505

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[groupthinking]

Duplicate of #1525, which covers the same three residual JSON-logging-fallback holes (raising __str__ on the exception, forged __class__, oversized ints/non-finite floats) and was already resolved via #1515. Closing as duplicate.

---

Generated by Claude Code

[groupthinking]

Duplicate of #1501, which is the currently-open canonical issue for the _format_json fallback still losing records — it already covers the oversized-int hole named here, and the same fallback has a long history of near-duplicate refilings (#1498, #1502, #1505, all closed as duplicates pointing back to this function). Closing this one in favor of #1501 so the "forged class values" / "non-finite floats" angles get tracked in one place rather than re-split. Please reopen if you believe this covers something #1501's acceptance criteria doesn't.

---

Generated by Claude Code