[CHA-4943] Retry once on stale keep-alive sockets

[itsmeadi]

Ticket

• https://linear.app/stream/issue/CHA-4943/ruby-sdk-pooled-tls-eof-against-gcp-load-balancers

Summary

Follow-up to #78 (idle_timeout 55→25). PrizePicks still hit GetStreamRuby::TransportError on reused keep-alive sockets (SSL_read EOF, Connection reset by peer, Net::ReadTimeout on a closed TCPSocket). Faraday/net_http_persistent does not retry those; Go net/http does for unused idle connections.

Retry the request once, with no backoff, when the failure looks like a dead pooled socket. Applies to POST as well as GET. DNS failures and real read timeouts are not retried. The opt-in retry_config: policy is unchanged.

Previous PR

• fix: close pooled TLS before GCP LB idle timeout [CHA-4943] #78

Checklist

• The changed code has been covered with unit tests
• API endpoints are covered with client tests
• The internal documentation (./docs) has been updated

Notes for review

• Multipart uploads are not retried (they skip Client#request before the retry loop).
• PrizePicks is still on 8.0.2; this needs a gem release after merge.

Made with Cursor