Claude Opus 5.5 (effort default (no --effort flag) · thinking on) on Fix a flaky test suite caused by set ordering, ignored date parameter and global random state - without touching the tests

[
 {
  "duration_api_ms": 17478,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.12006820000000001,
  "usage": {
   "input_tokens": 8,
   "cache_creation_input_tokens": 8568,
   "cache_read_input_tokens": 69261,
   "output_tokens": 1882,
   "output_tokens_details": {
    "thinking_tokens": 266
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 8568,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 496,
     "cache_read_input_tokens": 20488,
     "cache_creation_input_tokens": 226,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 226
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-opus-5-5": {
    "inputTokens": 8,
    "outputTokens": 1882,
    "cacheReadInputTokens": 69261,
    "cacheCreationInputTokens": 8568,
    "webSearchRequests": 0,
    "costUSD": 0.12006820000000001,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 266,
    "canonicalModel": "claude-opus-5-5",
    "provider": "firstParty",
    "costBasis": "list"
   }
  },
  "permission_denials": [],
  "terminal_reason": "completed",
  "fast_mode_state": "off",
  "fast_mode_disabled_reason": "sdk_opt_in_required",
  "subagent_stats": {
   "spawned": 0,
   "requested": {
    "background": 0,
    "foreground": 0,
    "unset": 0
   },
   "started_in_background": 0,
   "max_depth": 0,
   "spawned_by_subagents": 0,
   "completed": 0,
   "failed": 0,
   "killed": {
    "parent": 0,
    "user": 0,
    "system": 0
   },
   "refused": {
    "depth_limit": 0,
    "concurrency_limit": 0,
    "budget": 0
   },
   "by_type": {}
  },
  "safety_stops": 0,
  "is_error": false,
  "num_turns": 5,
  "subtype": "success",
  "api_error_status": null,
  "result": "I rewrote `billing.py` and all three tests now pass under five different `PYTHONHASHSEED` values (0, 1, 7, 42 and random). pytest isn't installed here, so I called the test functions directly. I didn't touch `tests/test_billing.py`.\n\nThere were three causes of the flakiness:\n\n1. **`today` was always overwritten** with the real current date, so the due-date and invoice-number tests only passed on certain days. It now defaults to the current date only when `today` isn't passed.\n2. **Customers came from an unsorted `set`**, so their order (and therefore invoice numbering) changed with the hash seed. They're now sorted case-insensitively, with ties broken by the exact name. The test's expected order is `acme, Bravo, delta, zeta`.\n3. **The audit sample used the shared global `random` module.** It now uses a private random generator seeded from a SHA-256 hash of the date, `audit_size` and each customer's name and total. The same inputs always give the same sample, whatever the hash seed or global random state, and the global `random` state is left untouched (I checked that it's identical before and after a call).\n\nThe sample still looks random and changes by month. For example, Jan\u2013Apr 2026 give samples of 003/004, 001/003, 004/003 and 003/002. The seed uses the full date, so two runs in the same month with different days will usually pick different invoices. I also now calculate the due date once per batch instead of once per invoice; the result is the same.",
  "ttft_ms": 1771,
  "type": "result",
  "duration_ms": 18050,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 779,
  "time_to_request_ms": 208,
  "first_content_frame_ms": 779,
  "queued_turn_count": 0,
  "result_index": 0
 }
]