test: Investigate How to Improve Test Reliability

[0xNeshi]

This issue is super important to handle to increase crate reliability and robustness.

In-memory Anvil node is very flaky when reorgs are involved. We should research whether using a Fully-featured Local Test Node Improves test reliability. If not, find a way to improve it.

[0xNeshi]

Also, take a look at the flaky live_mode_processes_all_blocks_respecting_block_confirmations test:

https://github.com/OpenZeppelin/Event-Scanner/actions/runs/18875721562/job/53864871746?pr=140

[0xNeshi]

Others have reported that Anvil has unreliable (to put it mildly) reorg behavior:
foundry-rs/foundry#10809

EDIT: this might even be an issue with alloy, but not sure yet:
alloy-rs/alloy#2668

[0xNeshi]

I managed to isolate the reorg bug and I created a detailed issue on alloy's repo:
alloy-rs/alloy#3116

Immediate measure: rewrite tests to avoid the reported edge case.
Long-term measure: investigate whether it makes sense to switch the local test node to Reth/Geth. alloy supports node binders for those, too: https://docs.rs/alloy-node-bindings/1.0.42/alloy_node_bindings/nodes/index.html

[0xNeshi]

Aside from the Alloy/anvil bug, it turned out the main problem was with the way we wrote our reorg tests. After refactoring them, and changing them to circumvent the alloy/anvil bug, the race conditions became obvious and were fixed. Time will tell if this really solved the problem.