bug: .env.sepolia uses identical bootnodes as .env.mainnet and inconsistent engine RPC scheme

[Sertug17]

Bug 1: Sepolia uses mainnet bootnodes

.env.mainnet and .env.sepolia contain identical BASE_NODE_P2P_BOOTNODES values. Sepolia testnet nodes bootstrap against mainnet peers, which causes inefficient peer discovery or failure to find valid Sepolia peers.

# .env.mainnet
BASE_NODE_P2P_BOOTNODES=enr:-J24QNz9lbrKbN4iSmmjtnr7SjUMk4zB7f1krHZcTZx-...

# .env.sepolia (identical should be different)
BASE_NODE_P2P_BOOTNODES=enr:-J24QNz9lbrKbN4iSmmjtnr7SjUMk4zB7f1krHZcTZx-...

Expected: Sepolia should have its own dedicated bootnode ENR records.

Bug 2: Engine RPC protocol inconsistency

.env.mainnet uses WebSocket:

BASE_NODE_L2_ENGINE_RPC=ws://execution:8551

.env.sepolia uses HTTP:

BASE_NODE_L2_ENGINE_RPC=http://execution:8551/

The consensus-entrypoint line 44 does a ws → http substitution for the health check, so the mainnet config works. But the inconsistency between environments is confusing and error-prone when operators copy configs between networks.

Impact

• Node operators running Sepolia may fail to discover peers or connect to wrong network peers
• Copying config between environments produces unexpected behavior due to protocol mismatch

[Sertug17]

I'd be happy to open a PR for both fixes. For Bug 1, could a maintainer share the correct Sepolia bootnode ENR records? The RPC scheme fix (Bug 2) is a one liner I can include in the same PR. @roethke