[aminsammara]

The staking summary's exiting and zombie counts and its TVL came from deposit and withdraw event arithmetic pooled across every rollup. This replaces that with getStatus reads per attester per rollup.

What was wrong

• zombieStakes was never counted. It was totalStakes − active − exiting, where totalStakes is deposits minus finalizes over all rollups. Any indexing gap, and every attester still validating on an older rollup, landed there.
• exitingStakes paired initiate and finalize events by attester across rollups.
• totalValueLocked was 200k × totalStakes minus slashes on the canonical rollup only. Attesters slashed on v5 that followed the latest rollup into v6 through the bonus instance kept their full 200k.

On mainnet the live API reported 3,632 total, 367 exiting and 106 zombie. Reading getStatus on all four rollups gives 3,533 total, 254 exiting and 70 zombie. Another 50 attesters are still validating on v5, and those ended up in the zombie count.

What it does now

• src/utils/attester-status.ts:
  • for every attester in the deposit table, and every rollup in rollup_version, calls getStatus;
  • then reads GSE.effectiveBalanceOf(rollup, attester) for VALIDATING attesters and getExit(attester).amount for EXITING and ZOMBIE ones;
  • reads in sequential multicalls of 1,000 calls. On mainnet that's 7,246 attesters × 4 rollups in about 16s;
  • caches the result for 60s. A stale snapshot is served at once while one background read refreshes it, and a failed refresh keeps the previous snapshot. Only the first request after startup waits.
• summary.ts builds exitingStakes, zombieStakes, totalValueLocked and activeValueLocked from that read.
  • activeStakes is still getActiveAttesterCount. It logs a warning if it disagrees with the canonical VALIDATING count from getStatus, which would mean the indexer is missing deposits.
  • New stats.legacyActiveStakes counts attesters validating on a non-canonical rollup.
  • New rollups[] gives a per-rollup breakdown.
• Both dashboard sublines show "+N on older rollups" when that bucket isn't empty.

The ABIs for getStatus, getExit and getGSE are identical on all four mainnet rollups (v4, v4.x, v5, v6), and they share one GSE. I checked each against live attesters.

Testing

• New unit tests for the reader and cache: yarn jest tests/unit/utils/attester-status.test.ts.
• I ran readRollupStakeTotals against mainnet with the full attester set. It matches a direct getStatus sweep.
• Indexer and dashboard typecheck clean, and the full indexer suite passes (39 tests).
• Fixed in this PR as well:
  • failed-deposits.test.ts was failing on main because the jest mock for ponder didn't export inArray.
  • ponder.config.ts failed typecheck with TS2742 because tsconfig.json asked for declaration output. Nothing emits declarations (the only tsc call is --noEmit), so I removed declaration and declarationMap.

Notes

• Multicall needs Multicall3 at its canonical address. Mainnet and Sepolia have it. A fresh local anvil without a fork doesn't, so the summary would fail there.
• Provider details and the beneficiary overview still classify attesters through attester-state.ts, which uses events. They can move to the same reader in a follow-up.