One ingest pipeline is degraded. Every incident below is written after the fact by the engineer who handled it, not summarised into nothing.
| Service | Last 90 days | Uptime |
|---|---|---|
| Sentinel — screening API | 99.98% | |
| CreditGraph — profile API | 99.96% | |
| PayRoute — quote & execute | 99.99% | |
| Ledger Mind — reconciliation | 99.97% | |
| Ethereum ingest | 99.99% | |
| Solana ingest | 99.61% | |
| Tron ingest | 99.98% | |
| Webhooks delivery | 99.95% | |
| Review console | 99.94% |
Uptime is measured per minute against a synthetic probe from three regions, and counts a response slower than five seconds as unavailable. Contractual credits apply on Enterprise plans below the committed figure.
A validator upgrade on our primary RPC provider is producing intermittent gaps in account subscription streams. Screening and credit responses on Solana are still served, and carry an as_of_block behind head — check that field rather than assuming freshness. Ethereum and Tron are unaffected. We are failing over to a secondary provider; next update within two hours.
A poison message in the delivery queue blocked the head of a shard, delaying webhooks for roughly 9% of customers by up to 41 minutes. No events were lost; all were delivered after the block cleared. We have since added per-message dead-lettering after three failures, which would have contained this to a single event.
A cache invalidation bug caused full re-tracing on repeat lookups of the same address, tripling load. p99 reached 6.4s; p50 was unaffected. Calls that exceeded five seconds were not billed. Fixed by keying the cache on block height instead of a timestamp window.
Two hours of TRC-20 transfer events were ingested late after a node fell out of sync without failing its health check. Scores computed in that window carried a stale as_of_block and were recomputed automatically. Health checks now assert head progression, not just liveness.
Incidents older than twelve months are archived and available on request. Illustrative history on this build — wire this page to your monitoring before launch.
Status changes push to Slack, PagerDuty or a webhook. Customers materially affected are also emailed directly — a status page is not a substitute for telling you.