# Week 13

Time for more benchmarks!

#### Latencies
Median (5m rolling):

Mean (5m):

p95:

median sits ≈12s and p95 stays ≲24–30s. That matches the spec target of 1–2 slots for confirmation; occasional bumps are expected near epoch boundaries as attestations roll in.

#### Slot Delays

Median slot delay (5m):

p95 slot delay (30m):

median ≈1 slot and p95 ≲2–2.5 slots. This is the slot-form view of the same latency and confirms most blocks meet the Q‑indicator in the next slot, with a small tail into the second slot.

#### Throughput

Confirmations per minute (1m rate):

Histogram-driven confirmations per slot (1m):

confirmations/minute and histogram‑driven confirmations/slot track the network’s block cadence (≈1 per slot during steady state). This shows FCR is neither under‑confirming nor congested.

#### Outcome Distribution

Zero-slot rate (should be ~0 - FCR isn’t confirming before there’s an attestation window due to our same-slot guard):

30m increase by bucket (stacked series):

Share of 1–2 slot confirmations (30m):

Share of ≥3 slot confirmations (30m):

1–2 slot share high, ≥3 slot share near zero: healthy network and correct thresholds. we can drop the same‑slot guard—at the cost of re‑introducing occasional 0‑slot artifacts.

#### Validator Support

Median validator support % for confirmed blocks (5m):

p95 validator support % (30m):

median and p95 support % are high (typically >80–90%), consistent with LMD‑GHOST votes converging quickly.

#### Safe head and reorg hygiene

Safe head slot (monotonic trend):

Reorg count rate (should be ≈0 in steady state):

Reorg depth p95 (if any):

reorg count rate ≈0 and reorg depth p95 ≈0: safe‑head stability is good, indicating the confirmed set is not oscillating.