fix: deflake should_flag_only_deposits_at_or_above_the_per_token_minimum

[basvandijk]

Problem

//rs/ethereum/cketh/minter:deposit_from_cex kept flaking after #11383 (which removed the dominant stop-canister race). Both post-#11383 master flakes share one new signature: should_flag_only_deposits_at_or_above_the_per_token_minimum fails at its first assertion with

assertion failed: `Scanning { valid_until: …, last_scanned_block: Some(Nat(11)), scan_count: 1 }`
does not match `DepositStatus::AwaitingSweep(detected) if …`

right after credit_deposits funded the deposit addresses on anvil (its own balance assertions passed, so the funds were readable).

Root cause

The minter's balance scan does not fetch a block number: it eth_calls at its cached latest_block_height (rs/ethereum/cketh/minter/src/balance_scan/mod.rs, pinned as BlockTag::Number), and that cache is refreshed only by the separate 30-second REFRESH_LATEST_BLOCK_HEIGHT_INTERVAL timer (deposit.rs is its sole writer; the log scrape writes a different field). So a scan firing after the harness mined the funding block can still read every balance at a pre-funding block: the call succeeds (the token code, set through anvil_setCode, is visible to historical eth_calls — mined transfers are not), finds nothing, and record_scan stores scan_count = 1 with the stale last_scanned_block.

The failing logs show exactly this: an ETH log scrape covering blocks 1..=11 (head ≥ 11 pre-funding), the funding mined after it, then one balance scan — scanned 3 (address, token) pair(s), found 0 candidate(s), 0 decode error(s), 0 call error(s) — pinned at block 11.

The harness' await_scan accepted any Scanning status with scan_count >= 1 as terminal, so it returned that transient verdict instead of driving on until a scan that could actually see the funds — and the test's AwaitingSweep assertion failed. (The next scan corrects the verdict within a couple of blocks; the predicate just never waited for it.)

Fix

await_scan now snapshots the anvil chain head at entry (the funding is always mined by then — every caller funds through credit_deposits first) and counts an address as scanned only once its status is:

• AwaitingSweep — terminal by construction (a detection only results from a scan that read a balance at or above the minimum, and a queued pair never returns to Scanning), or
• Scanning with scan_count >= 1 and last_scanned_block at or past that entry head — a scan pinned there cannot have missed the balance, so a below-minimum verdict is genuine.

SCAN_TICKS grows from 4 to 8 (a budget, not a cost — driving stops the moment the condition holds): the pair's block-based backoff can demand up to 300 block-seconds before it is due for the scan the new predicate waits for, against the ~10 blocks each tick's settle mines.

Validation

• Deterministic reproduction: temporarily deploying the token code before registration (so a pre-funding scan completes, recording exactly the state the stale-cache race records — scan_count = 1 at a block where the funding is invisible) and forcing one such scan:
  • old predicate → fails with the exact CI signature, deterministically;
  • fixed predicate → converges to AwaitingSweep and the test passes end to end.
• cargo check, cargo fmt, ./ci/scripts/rust-lint.sh — clean.
• Depth-2 bazel rdeps all pass: integration_tests_tests/{ckerc20,cketh}_test, minter:lib_tests, test_utils:lib_tests.
• bazel test --runs_per_test=3 --jobs=3 //rs/ethereum/cketh/minter:deposit_from_cex — 3/3 full-suite runs pass.

The remaining non-success runs in the window since #11383 merged all ran on commits that predate it (branch/merge-queue checkouts from before the merge) and show the old, already-fixed stop-canister signature.

This PR was created following the steps in .claude/skills/fix-flaky-tests/SKILL.md.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Deflakes the CEX deposit test by ensuring balance scans include the funded block.

Changes:

• Requires scans to reach the Anvil head captured when waiting begins.
• Increases the scan-tick budget to accommodate backoff.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[zeropath-ai]

✅ No security or compliance issues detected. Reviewed everything up to 04e0051.

Security Overview

• 🔎 Scanned files: 1 changed file(s)
• 🔗 Scan Link: https://zeropath.com/app/repositories/783a9bd4-fb4e-4e10-ab3c-c2fb2ddfb92f?scanId=8f282240-9902-4160-a58a-df1df80bbb16&codeScanTypes=PrScan&tab=issues

Detected Code Changes

Change Type	Relevant files
Enhancement	► rs/ethereum/cketh/test_utils/src/live.rs       Modify SCAN_TICKS constant and augment await_scan doc comment and logic to account for block visibility and pinned-block requirements in balance scans ► rs/ethereum/cketh/test_utils/src/live.rs       Update await_scan implementation to consider last_scanned_block and funded_by conditions

[mbjorkqvist]

Thanks @basvandijk!