[BUG] eth_call discrepancy with public endpoint

[qn-srikanth]

Seid version

name: sei
server_name: <appd>
version: v6.0.2
commit: f17ec67f1a05c0fa1c3a978bad456edb362aa7c9
build_tags: netgo,
go: go version go1.21.13 linux/amd64
build_deps:
- cosmossdk.io/[email protected]
- filippo.io/[email protected]
- github.com/99designs/[email protected]

Chain ID
Mainnet / Pacific

Describe the bug
Inconsistent eth_call response with public endpoint

To Reproduce

Contract : 0x9851d050a3db9388628cd3da22944a10248a4f3b

{
  "jsonrpc": "2.0",
  "method": "eth_call",
  "params": [
    {
      "to": "0x9851d050a3db9388628cd3da22944a10248a4f3b",
      "data": "0xc87b56dd0000000000000000000000000000000000000000000000000000000000000296"
    },
    "latest"
  ],
  "id": 1
}

This is the Token that’s returning execution reverted when getting the URI but works with the public endpoint https://evm-rpc.sei-apis.com/ https://opensea.io/assets/sei/0x9851d050a3db9388628cd3da22944a10248a4f3b/662

Expected behavior
Return the token URI

Screenshots

Additional context
https://sei-evm-rpc.publicnode.com/ and https://1329.rpc.thirdweb.com/ have the same -32000 response.

[dvli2007]

Is there some query gas cap configured on different RPCs? I believe the sei-apis.com RPC is configured with 10M gas cap for queries.

This is a call to an ERC721 pointer contract, and the base contract is CW721. What's notable is that the CW721 metadata is fully onchain, the nft_info query on the CW721 returns >30KB of data, base64 encoded.

When querying the ERC721 tokenURI call, there's a staticcall made which seems to be significantly gas intensive. Furthermore, since the staticcall for nft_info returns more data than is needed by tokenURI, I imagine there is extraneous data that gets loaded into memory, contributing to the high gas usage.