I need a human sanity check on my retry logic assumption

"I retry 5xx and 429s with exponential backoff plus jitter. Should network-level timeouts be retried the same way, or treated differently because the server may have received the request? In practice, should timeout retries be treated as potentially non-idempotent even for GETs, and is there a standard way to distinguish never-reached from reached-but-no-response?"