Fast Confirmation Rule on Mainnet

Networks:
Ethereum Mainnet
Time range:
Start 2025-11-07T00:00:00Z
End 2026-03-17T23:59:59Z
consensus fast-confirmation attestations
Published March 24, 2026

Question

Would the Fast Confirmation Rule work on Ethereum mainnet?

Background

The Fast Confirmation Rule (FCR) is a way for consensus clients to locally confirm blocks within seconds of the attestation deadline, rather than waiting ~13 minutes for FFG finality. No hard fork needed, it's a client-side check.

The full algorithm operates on the fork choice store: it starts from a safe checkpoint, computes LMD-GHOST scores using each validator's vote and the committee shuffling for the current and previous epochs, discounts proposer boost, and iteratively advances a confirmed_root through the canonical chain. A block is confirmed when its support is strong enough that no adversary (up to β fraction of stake) could cause a reorg. If a block doesn't get enough support from its own slot's committee, it can still confirm as attestations from subsequent slots accumulate.

Investigation

All results here are estimates. Rather than replaying the full FCR algorithm (which evaluates the entire fork choice tree), we use per-slot head vote ratios as a proxy. The real algorithm computes a dynamic safety threshold based on committee weights across multiple slots, proposer boost, equivocations, and empty slot discounts. In the simplest case (block confirmed within its own slot, `CONFIRMATION_BYZANTINE_THRESHOLD = 25`), this simplifies to roughly 95% of the slot's committee voting for the head. We use 95% as our proxy threshold. We also use finalized attestation data rather than real-time observations, so attestation arrival timing, propagation delays, and local node views are not captured.

How many slots would confirm?

1,074,543 slots from Nov 7 2025 to Mar 17 2026 (131 days). Source: fct_attestation_correctness_canonical on xatu-cbt.

Per slot, we check what fraction of the committee voted for the head block (votes_head / votes_max). If it clears the threshold, that slot would've been fast confirmed.

View Query: fcr_head_vote_daily
Loading...

At the 95% threshold, 96.9% of slots would fast confirm. The worst day was December 4 during Fusaka-related network disruption, but even then 80% of slots still cleared the bar. Drop to a 90% threshold and the overall rate hits 98.4%.

Threshold Overall Worst Day (Dec 4)
>= 95% (β = 0.25) 96.9% 80.4%
>= 90% 98.4% 95.1%
>= 85% 98.8% 97.9%
>= 80% 99.1% 99.0%

The ~3% that miss the 95% threshold are slots with late blocks, missed proposals, or low participation.

Would any reorged block be confirmed?

1,900 orphaned blocks from Nov 7 2025 to Mar 17 2026. Source: fct_block_proposer FINAL joined with int_attestation_attested_canonical on xatu-cbt.

For every orphaned block, we check how much of the committee actually voted for it by matching each validator's beacon_block_root vote against the orphaned block's root.

No. None came close:

Metric Value
Orphaned blocks checked 1,900
Would confirm at >= 95% 0
Would confirm at >= 80% 0
Would confirm at >= 50% 0
Max support any orphan received 47.8%
Average support 2.1%
Median support 0.4%
Not seen by committee at all 20%

Reorged blocks are blocks the committee never saw. The median orphan had 0.4% support. FCR wouldn't touch any of them.

How fast would blocks confirm?

372 slots sampled from Mar 3-17 2026, observed by 46 sentry nodes across 5 cities. Source: libp2p_gossipsub_beacon_attestation and libp2p_gossipsub_aggregate_and_proof on the xatu cluster.

For each slot, we decode aggregation bitfields from gossipsub messages and track cumulative unique validator support over time to find when the 95% threshold is crossed.

Loading...

In this sample, 97% of slots confirm with a median time of 8,127ms after slot start, about 4.1 seconds after the attestation deadline. Geographic spread is tight: Helsinki confirms at ~8.1s, Sydney at ~8.25s, less than 200ms apart.

Note that this data only derives attestations from aggregates. By default, nodes would also observe 2/64 of the committee via unaggregated attestations, resulting in faster confirmations. Nodes that are manually subscribed to all subnets would most likely confirm even faster.

Takeaways

  • 96.9% of slots would fast confirm at the 95% threshold across 131 days and ~1M slots, hitting 97%+ on most days
  • Zero reorged blocks would be confirmed: the highest support any orphaned block received was 47.8%, with a median of 0.4%
  • Blocks confirm ~4 seconds after the attestation deadline in a 372-slot sample, with under 200ms variation across 5 cities