feat(funding): fund the account pool from a root key for real-chain runs

[bdchatham]

Problem

seiload generates random EVM accounts and never funds them, so today it only works against mock chains (mock_balances auto-tops EVM balances at execution) or a fresh genesis that pre-funds them. Pointed at a real, long-running chain (e.g. arctic-1), every generated account is at zero balance and every transfer fails.

What

Add an optional funding config block. Given one funded root key, seiload funds its generated account pool via the existing Disperse contract at startup — before prewarm and dispatch — so it can drive load on a real chain. Absent the block, behavior is byte-for-byte unchanged (mock/genesis path).

"funding": {
  "rootKeyEnv": "SEILOAD_ROOT_KEY",   // hex ECDSA key from env (preferred; keeps it out of the file/logs)
  "fundAmountWei": "1000000000000000000",  // per-account, default 1 SEI
  "batchSize": 200,                    // recipients per disperseEther call
  "disperseAddress": "0x…"             // optional: reuse a pre-deployed Disperse
}

How

• config (funding.go): FundingConfig; a BigInt string-JSON wrapper (wei exceeds 2^53 so JSON numbers would lose precision); ValidateFunding() rejects a missing key and newAccountRate>0 under funding (on-demand accounts can't be funded → their first tx would fail for gas).
• funder (funder/funder.go): parse root key → dial → enumerate the pool → skip already-funded accounts (idempotent across pod restarts) → deploy or reuse Disperse → batch disperseEther. The root's first EVM tx auto-associates its cosmos balance to the EVM side (Sei ante handler), so no explicit association step.
• types: AccountPool.GetAccounts() to enumerate the fixed pool.
• main: funding runs after sender init, before prewarm; no-op when funding is nil or in dry-run.
• profiles/arctic-1.json: example real-chain profile (note chainId: 713715 — arctic-1's EVM chain id, not the 713714 local default).

Test

• config/funding_test.go: BigInt string round-trip (incl. >2^53 + rejecting bare JSON numbers), funding defaults, and ValidateFunding cases.
• go build ./..., go vet, go test ./config/... ./types/... ./generator/... all green; gofmt clean.

Why (context)

Restores the synthetic load that kept arctic-1 at ~19 TPS (it stopped when the migration disarmed the EC2 node that ran the old co-located loadtest), and lets us run a load generator per region to study us-east-2 commit participation. Funded from recovered genesis test accounts (funny money); the key ships as one SOPS secret per region on the platform side.

🤖 Generated with Claude Code

[cursor]

PR Summary

High Risk
Loads a root private key and sends on-chain value to many accounts; misconfiguration or compromise could drain the funder wallet despite validation and dry-run skipping.

Overview
Adds optional funding config so seiload can fund its fixed account pool from one root key before prewarm/dispatch, enabling load against real chains where generated accounts start at zero.

Config introduces FundingConfig (rootKeyFile preferred over rootKeyEnv, string-JSON fundAmountWei, batch size defaults), ValidateFunding() (requires a key; rejects newAccountRate > 0 on top-level or scenario pools), and wires validation into config load.

Runtime adds the funder package: resolve key → dial RPC → dedupe pool addresses via new AccountPool.GetAccounts() → skip already-funded balances → deploy and verify a fresh Disperse contract → sequential batched disperseEther with mined receipt checks. main calls funding after sender setup when funding is set and not dry-run; nil funding leaves behavior unchanged.

Includes profiles/arctic-1.json as a real-chain example and config/funding_test.go for BigInt JSON and validation cases.

^{Reviewed by Cursor Bugbot for commit 414ec7c. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit b655eeb. Configure here.}

[bdchatham]

Both Cursor Bugbot inline findings were flagged against the first commit and are already resolved as of d1fe201 (independently corroborated by a specialist cross-review — they found the same two):

• "Ignored reverted funding transactions" (High) → fixed in 7945b34: added waitSuccess(), which asserts receipt.Status == ethtypes.ReceiptStatusSuccessful after the Disperse deploy and every disperseEther batch (funder.go:222). A reverted tx now aborts the run instead of proceeding silently. Deploy also gets a CodeAt-non-empty check.
• "Balance check abort leaks workers" (Medium) → fixed in 7945b34: filterUnderfunded was rewritten on errgroup.WithContext + SetLimit(16); the hand-rolled unbuffered channels are gone, so the early-error/ctx-cancel paths cancel in-flight queries and join cleanly with no goroutine leak.

Also addressed from the same review: dropped the unvalidated disperseAddress (always self-deploy now), removed the inline rootKey field (rootKeyFile preferred / rootKeyEnv fallback), and corrected the idempotency doc (fresh random pool per restart). CI green (build-and-test 1.25.1, container build, Bugbot).

[amir-deris]

Does this file need to be a README instead? it seems it doesn't have any code 🤔