… prior simulated blocks

simulationIntraBlockStateReader.HasStorage only queried RangeAsOf(firstMinTxNum),
missing storage written to the in-memory RAM batch by prior simulated blocks.
This caused false negatives for newly deployed contracts, breaking EIP-7610
CREATE2 collision detection in multi-block simulations.

Fix: add HasPrefixInRAM to TemporalMemBatch interface and implement it as a
btree-only scan (no disk I/O) for StorageDomain. HasStorage now checks the
RAM batch first before falling back to RangeAsOf.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Override called FinalizeTx(rules, NoopWriter) to clear the journal after
applying state overrides. For any account that became empty (nonce=0, code=0x,
balance=0) after the override, FinalizeTx triggered EIP-161 empty-account
removal: DeleteAccount was discarded by NoopWriter but stateObject.deleted was
set to true. This caused IntraBlockState.HasStorage to short-circuit to false
before reaching the state reader, breaking EIP-7610 CREATE2 collision detection
in multi-block eth_simulateV1.

Fix: use SoftFinalise instead of FinalizeTx(NoopWriter). SoftFinalise clears
the journal (preventing revert from undoing overrides) without triggering
EIP-161 account deletion.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Override called FinalizeTx(rules, NoopWriter) to flush dirty storage into
originStorage (needed for correct SSTORE gas). However, EIP-161 empty-account
removal also fired: any account that became empty (nonce=0, code=0x, balance=0)
after the override was marked deleted=true in the IBS, even though the deletion
was discarded by NoopWriter. This caused IntraBlockState.HasStorage to
short-circuit to false before reaching the state reader, breaking EIP-7610
CREATE2 collision detection in multi-block eth_simulateV1 when a prior
simulated block had deployed a contract at the overridden address.

Fix: copy rules and set IsSpuriousDragon=false before passing to FinalizeTx.
This preserves the originStorage commit (gas correctness) while preventing
EIP-161 from deleting accounts during stateOverride processing. State overrides
are simulation-only mutations and must not trigger consensus rules.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

… accounts

In multi-block eth_simulateV1 simulations, stateOverrides.Override calls
FinalizeTx which marks overridden accounts as dirty in stateObjectsDirty.
CommitBlock's MakeWriteSet then iterates all stateObjects and applies EIP-161
to every dirty account — so an override-only account with nonce=0/code=empty
is spuriously deleted, which also wipes its storage via DomainDelPrefix.

Fix: after Override, snapshot and clear stateObjectsDirty
(ExtractAndClearDirty) so CommitBlock only sees transaction-touched accounts.
After CommitBlock, CommitOverrideDirtyAccounts writes the override-only
accounts with EIP-161 disabled, preserving their state in the simulation.

This fixes the stateRoot mismatch in test_28 where a CREATE2-deployed
contract's storage survived across blocks as a collision target.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- ExtractAndClearDirty: use maps.Clone instead of manual copy loop
- CommitOverrideDirtyAccounts: pass false directly to updateAccount instead
  of going through a noEIP161Rules intermediate copy
- HasPrefixInRAM non-StorageDomain path: use strings.HasPrefix to avoid
  []byte(k) allocation per entry

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>