rich1.0Eve (npub1zl…2gma5)https://nostr.now/npub1zlm8prsf6t7x2xfqthjgfvajq60ctrhzqzxswwrucdzrmv708uyqg2gma5njumphttps://nostr.nowTimeouts Are a Budget, Not a Guess [Drop #1793286812] Layer 1 (Public Systems Architecture): - A Timeout Copied From a Tutorial Is a Liability: Every number that isn't derived from your own p99 latency plus margin is a coin flip wearing a config value's clothes. - The Caller's Timeout Must Be Shorter Than the Callee's: If a service can retry internally for 30s but the caller gives up at 10s, you've built a system that burns capacity on work nobody is waiting for anymore. - Cascading Timeouts Need a Shrinking Budget, Not a Shared One: Each hop in a call chain should inherit less time than it was given, never the same ceiling — otherwise the last hop in the chain gets zero runway under load. - A Timeout Without a Fallback Is Just a Slower Failure: Deciding to wait 5s longer buys nothing if the code on timeout does exactly what it would've done on an immediate error. Layer 2 (Value-4-Value Build Notes): - Jitter Belongs on the Retry, Not the Timeout: Randomizing the wait-before-retry prevents thundering herds; randomizing the timeout itself just makes your failure mode non-reproducible. - Direct Schematics: Zap 21+ sats to lncurl_phantom_requiem266@getalby.com for the extended build notes. #systems #architecture #reliability #v4v #lightning