The full report’s core finding: customers do not buy “RPC” as a neutral endpoint. They buy fewer production unknowns. Competitors can keep claiming speed, uptime and chain coverage; GetBlock’s sharper wedge is to make operational risk measurable, priced and contained.
Executive findings
- Toxic workloads are the strongest attack surface.
eth_getLogs, archive access, polling, websocket subscriptions and backfills create disproportionate infra cost and customer pain. - Failure containment beats generic latency claims. Buyers need method-level status, RCA, endpoint diagnostics and degradation handling.
- Multichain breadth is not enough. Chain count must become a control plane: scoped keys, rotation, per-chain/method quotas, anomaly alerts, invoice/VAT hygiene and overage guardrails.
- L2/appchain/agent growth is a timing play. Dencun supports the L2 trajectory, but RaaS/indexing/agent platforms may bundle away RPC procurement.
Evidence discipline
Vendor pages and internal competitive-intel pages are treated as interested sources. They are useful for positioning and product-direction signals, not independent proof. The stronger operational evidence layer comes from GitHub artifacts showing rate limits, degraded endpoints, credit-limit changes and weighted/CU rate limiting in real software contexts.
Site implication
The website should now treat the daily monitor as input, not output. Every new competitor change should be routed into one of three decision tracks: Cost Profiler / Toxic Workload Suite, Failure-containment layer, or Appchain/agent migration layer.