RPC behavior and rate limits

[asseph]

• Using rpc.MainnetRPCEndpoint directly can easily hit HTTP 429 “Too many requests,” especially with heavy methods like getTokenAccountsByOwner, which may surprise users who assume the error is from the SDK and not from the public endpoint limits
• Switching to a third‑party RPC often requires changing endpoints and understanding that some providers have custom APIs or behaviors; if their JSON-RPC behavior diverges from Solana’s official spec, the SDK may behave unexpectedly

[luanxu-dev]

I ran into this exact confusion before. A practical mitigation is to treat rpc.MainnetRPCEndpoint as best-effort only, and route heavy calls (getTokenAccountsByOwner, large scans) to a dedicated endpoint with method-level throttling. I also separate 429 handling from generic retries (short jitter backoff + fallback endpoint) so SDK users can see this is endpoint capacity, not SDK correctness.