The strongest conclusion in the attack-surface report is that generic RPC/CU pricing is fragile. The valuable workloads are not cheap eth_call. They are heavy, bursty, stateful and hard to forecast: eth_getLogs, archive reads, traces/debug, websocket subscriptions, polling loops, indexer backfills and historical scans.
Product proposal
- RPC Cost Profiler: top methods by CU, request count, error rate, latency p95/p99 and support-ticket correlation.
- Logs/Archive/Backfill SLO: explicit ranges, latency/error targets, cache/indexing alternatives and backfill windows.
- Websocket stability plan: subscription limits, reconnect behavior, alerting and workload isolation.
- Stream/webhook alternative: migration path away from repeated polling when raw RPC is the wrong data-delivery primitive.
Fast validation
Two-week beta: export top methods by CU/error/latency, identify 20 heavy-workload customers, send each a read-only Cost Profiler report, offer a beta Logs/Backfill SLO, and measure willingness to pay, 429/support-ticket reduction and gross margin by method.
Sales angle
“Bring us your current traffic mix. We will show which methods create cost, latency and incident risk — then map it to a predictable production setup.”