fix(yearn): retry Envio requests with a 10s per-attempt timeout

[spalen0]

Why

yearn-alert-small-parent-flows sent Envio GraphQL request failed (HTTP Error 504: Gateway Timeout) alerts on 4 of 24 hourly runs (Sep 26 21:05 and 22:05, Sep 27 17:05 and 18:05 UTC). The hosted Envio indexer returned a gateway timeout after 18–36s, and a single failed request skipped the whole run. When I re-ran the same queries with the production saved positions, they finished in 0.1–1.4s, so these are short upstream slowdowns, not a query problem.

What changed

• alert_small_parent_flows.py and alert_large_flows.py now send Envio requests through utils.http_client.request_with_retry. It retries 5xx errors, 429s, timeouts and connection errors with exponential backoff. The retry work in fix(http): add retry logic to remaining HTTP requests #381 missed these two files because they used urllib directly.
• Each attempt now times out after 10s instead of 30s, so a fully failed run gives up after about 47s.
• Small-flow alerts show a short reason (HTTP 504 Gateway Timeout, timed out after 10s) rather than the raw exception, which would include the Envio endpoint URL.
• Large-flow alerts now also cover timeouts and connection errors. Before, those crashed the script.

Tests

• Added tests for retrying a 504 with the 10s timeout, and for a single alert without the URL when every retry fails.
• pytest: 1401 passed, 6 skipped. ruff check is clean. mypy reports no new errors.
• Ran a real query through both scripts against the production Envio endpoint and both got data back.

🤖 Generated with Claude Code