Query dispute: ftnGO (QmRmv5…) on mainnet, staked.cloud missing a Filled record

Filed a conflicting-attestation query dispute under GIP-0085 §3.2 on the ftnGO subgraph, QmRmv5cDAFtqs53NbtybGpaTqMgR2RPvpGWkDWc8Tk6UWG on mainnet. Dispute 0x15afb3a5e0ac673589652daeeb43a59eaee740d711f81eaf934eacc0bd6005ab is against staked.cloud 0xdc53e62df89fd07b31ed4ff886397b9e7ae4625e, dispute 0xb488f6d7b11c588970da1980ec888fb42c075c3f8be70a3e37efe529154fb969 is the other half against 0x8bbe94c2894f76406568dfb44e905dac4b7df699, tx 0x63e71ec665246245721b57ec276009915dc1b065052d7ac3eb7773d954e4bbc2.

The same block-pinned query got two signed answers. Three staked.cloud allocations return a filled list that is missing one record. The other indexer returns it. The record is a mainnet event in tx 0x0f91931debdf99de7ed468b7b4e1ee90a9f64b9adab8756f0a653c50d4df0f72, block 21,804,357, February 2025, and an independent graph-node sync from genesis holds it. Full report, attestations and the reproduction steps: Query conflict: a missing Filled record from staked.cloud on the ftnGO subgraph (mainnet) - HackMD

Same shape on governance-v3, QmRyQo7xz4P3WvthdVzfkhwSzfwa8Uku1qrcyJCAN6RAWS, from the same nodes, reproduced the same way. It is in the report as context, not filed.

Happy to answer anything the arbitrators or Staked want checked.

Hi @AxiomaticAardvark and arbitrators,

We’ve reproduced it and agree with the finding.

**What we confirmed**

- Our response to the disputed query hashes to the attested `responseCID`, so it is the answer our nodes served, and they still return it.

- Our ftnGO store has no `Filled` entity for tx `0x0f91931d…` (block 21,804,357). Comparing against the contract’s on-chain logs, we’re missing 11 events in total, in three block windows: about 21,088,000–21,093,500 and 21,116,000–21,122,900 (November 2024), and 21,804,300–21,804,600 (February 2025). Events on either side of each window are present.

- governance-v3 has the same problem: `constants` id `100` doesn’t exist in our store. That deployment keeps full history, so pruning isn’t the cause.

- All staked.cloud indexers query the same graph-node store, which is why the three allocations returned identical responses.

**Likely cause**

The gaps look like `eth_getLogs` responses that came back successful but incomplete during the original indexing. graph-node treats a successful response as complete, so it skipped those events without an error. Our current store is a January 2026 copy of the original deployment and inherited the gaps. We no longer have RPC or graph-node logs from that period, so we can’t tell which provider returned the incomplete data.

**What we’ve done**

On October 21 we stopped allocating to both deployments and closed every staked.cloud allocation on them; none remain open. We intended to close them with a zero POI, but our close command omitted `–force`. The agent rejected the zero POI, and the allocations were then closed with the POI graph-node computed from the incomplete store. We didn’t intend to further profit off of incorrect data and we’re going to hold off allocating to these subgraphs until this dispute is resolved and we’ve successfully resynced. We’re correcting our runbook so future zero-POI closes are submitted as intended.

**Questions**

1. Is there any additional information you’d like from us?

2. We’d like to start resyncing both subgraphs. Do you need us to keep the current data for further proof of indexing? If so, would a database dump of both deployments, kept until the dispute is resolved, be enough, or would you prefer we keep the current store live and queryable?

3. Before we allocate again, we plan to verify the resynced data against on-chain logs and other indexers’ POIs. Are there other steps you’d recommend to make sure this doesn’t happen again?

Thanks,

Staked Team

Note this response was generated with some AI help :slight_smile:

Thanks nate, that matches what we see on both deployments, and the three windows are a more precise account of the gap than I had.

  1. No. The report plus your confirmation is what the arbitrators need from either of us.

  2. Keep the current store live and queryable until they rule if you can, a dump with the table hashes is the fallback. We have independent syncs of both deployments already, so nothing on our side depends on it.

  3. Before you allocate again, compare your public PoI at a shared epoch block with two other operators, from a node that still holds that block. The 21 October closes went out with the computed PoI rather than zero, that is on chain and for the arbitrators to weigh, not me.

Hi, we are reviewing the information and will come back shortly.

In the meantime here is the dispute information:

Dispute (0xb488f6d7b11c588970da1980ec888fb42c075c3f8be70a3e37efe529154fb969)
├─ Type: Conflicting
├─ Status: Undecided (7.91 days ago)
├─ Indexer: 0x8bbe94c2894f76406568dfb44e905dac4b7df699
├─ Fisherman: 0x083a086ca2ccec6526cbe97b2841a481f99834af
├─ SubgraphDeployment
│  └─ id: 0x330bf5f206e467734b5ff7cff88cee131c05b96b717390a3bce5d337d9f7c465 (QmRmv5cDAFtqs53NbtybGpaTqMgR2RPvpGWkDWc8Tk6UWG)
└─ Economics
   ├─ indexerSlashableStake: 424309.320726609653676072 GRT
   └─ indexingRewardsCollected: 0 GRT
Dispute (0x15afb3a5e0ac673589652daeeb43a59eaee740d711f81eaf934eacc0bd6005ab)
├─ Type: Conflicting
├─ Status: Undecided (7.91 days ago)
├─ Indexer: 0xdc53e62df89fd07b31ed4ff886397b9e7ae4625e
├─ Fisherman: 0x083a086ca2ccec6526cbe97b2841a481f99834af
├─ SubgraphDeployment
│  └─ id: 0x330bf5f206e467734b5ff7cff88cee131c05b96b717390a3bce5d337d9f7c465 (QmRmv5cDAFtqs53NbtybGpaTqMgR2RPvpGWkDWc8Tk6UWG)
└─ Economics
   ├─ indexerSlashableStake: 1893699.193391486530929781 GRT
   └─ indexingRewardsCollected: 0 GRT

Just wondering, any update on this? We’d like to start resyncing these subgraphs but don’t want to delete the data before the dispute is fully resolved.

Thanks,

Staked Team