The v3.3 changes are in the repo, along with a new report.md and updated launch steps. Nothing was deployed. forge build, forge test (159 passed, 0 failed, 1 opt-in fork test skipped) and forge fmt --check all pass with solc 0.8.26. The brief asked for a second contributor with fresh context to review, and that did not happen: no other contributor was available in this run. I did my own checklist review instead (docs/review-v33.md) and fixed what it found, so an independent review is still needed before launch.

What changed

1. Idle farms stop diluting: a farm leaves the global hash total 12 epochs (3 days) after its last owner action. advanceEpoch does this with one bucket per epoch, and a test shows 15 farms expiring together cost the same as one.
   • Rewards earned before expiry are kept: a farm settled late gets exactly what it would have got on time.
   • Sync, claim, install, repair and pack re-commit the farm from the next epoch. There is a new FarmCore.sync() for this.
   • Claim only re-commits once the farm is fully caught up.
2. Anti-rush:
   • Level caps by farm age: level 5 needs 1 day, level 7 needs 3 days, level 10 needs 5 days. A test shows paid XP can't get around them.
   • Paid XP is capped at 400 per farm per UTC day; activity XP has no cap.
   • Merging is removed.
   • The weekly best district is scored as average spend per active farm, with no wallet counting for more than 10% of the district score.
3. Team share: packs, lobbying and the sell surcharge split 88% reward pool / 10% team / 2% heartbeat (1 ETH cap, excess to the pool). The swap base fee stays 1% / 1%.
4. Launch protection on the hook (fixed in the bytecode, no owner):
   • The buy tax starts at 90% and falls linearly to the normal 2.25% over 30 minutes. Everything above the normal fee goes to the reward pool.
   • Each buy is limited to 0.5 ETH for the first 60 minutes, including the FeeRouter's own buybacks.
   • The site can read launchStatus(), currentBuyTaxBps(), launchTaxTimeLeft(), maxBuy() and maxBuyTimeLeft().
   • PackSale opens at token launch + 60 minutes. Its price curve starts counting at opening, not at launch.

Simulation (20% endowment, 42-day drip, players sync once a day)

• Team ETH is lower than v3.2 because the team share dropped from 13% to 10%.
• Drawdown is higher than v3.2 because more tokens reach active players, who then sell.
• Whale (300 packs on day 1, 5.94 ETH): its emission share is 4.04% on day 1, 3.18% on day 3 and 2.11% on day 7.
• Snipers (5 ETH in the first minute as ten 0.5 ETH buys):
  • They get 48M tokens, against 270M with no launch tax.
  • 4.31 ETH of tax goes to the reward pool.
  • The price rises 15.3%, against 165% with no launch tax. With no other trading modelled, it is the same at 30 and 60 minutes.
  • Dumping loses them about 4.5 ETH.

Other changes you should know about

• Several existing tests were adjusted to the new rules, and a test helper fakes an older, higher-level farm for tests that need legendary or mythic items.
• I fixed a clock test that was already failing in v3.2 under the installed Foundry version.
• Module bytecode pins and the flattened launch/V4Market.sol were regenerated.
• The old report moved to docs/report-v32.md.

Decisions for you

• Is the 10% wallet cap too strict? Read literally, it means a district with fewer than 10 spending wallets scores zero. A softer rule (no wallet counts for more than 10% of its district's total spend) lets small districts compete and is a one-line change.
• **Launch timestamp