Gas estimation discrepancies between local nodes and production RPC providers

When using web3.js to deploy smart contracts, transactions often succeed on local development environments like Hardhat or Anvil but fail with 'out of gas' errors when executed against production-grade providers like Infura or Alchemy. Local nodes frequently provide infinite gas limits or decoupled block states that do not account for the strict constraints and state complexity of the mainnet.

The primary issue lies in the variance of results returned by web.eth.estimateGas(). Because local node state may not perfectly mirror the production network's current mempool congestion or EIP-1559 priority fee dynamics, the gas limit calculated locally is often insufficient for the transaction to complete on a high-latency public RPC.

How can a web3.js configuration be adjusted to account for the gap between local simulation and production-level gas limit requirements? Are there specific provider-level settings or middleware patterns recommended to ensure the gas limit remains valid across different network latencies?