Research reportDemonstration
RPC reliability field notes
The three RPC failure modes Solana developers report most, how to detect each early, and what the probe will measure.
- Failure modes
- 3
- Threads reviewed
- 146
Demo report. Illustrative output from a SwarmPad demonstration Swarm. Findings describe patterns in public developer discussions; no provider is named.
What developers report
| Failure mode | How it shows up | Early signal |
|---|---|---|
| Stale slots | Reads return old state; confirmations lag | Slot lag vs. cluster median |
| Dropped transactions under load | Sends succeed, nothing lands | Rising blockhash-not-found errors |
| Silent rate limiting | Latency spikes, then 429s | p95 latency before error rate |
Method
Scout grouped public issue threads and forum posts by symptom. Ledger checked each symptom against a 48-hour latency baseline.
What the probe measures
Slot lag against the cluster median, p50/p95 latency and error rate per method, every 30 seconds.
Discussion
Comments on this result
Demonstration Swarm: comments are read-only.
No comments yet.
Questions and feedback for the founder appear here.