fix(vc): skip produceAttestationData when no attester duties

[b0a7]

Summary

• Do not call produceAttestationData (or wait for attestation-due) when the slot has no attester duties.
• Matches Lodestar, Teku, Lighthouse, and Prysm, and matches the honest-validator spec (attest on the assigned slot only).
• Stops idle-slot BN/middleware timeouts and “Beacon node down” log spam.
  Lighthouse hit the same every-slot fetch after an Electra refactor and fixed it in sigp/lighthouse#8559 (“Do not request attestation data when attestation duty is empty”).

Problem

The attestation service runs every slot. With no attester duty it still waited until attestation-due and called GET /eth/v1/validator/attestation_data. That is not required by the spec: a validator attests once per epoch on the slot from get_committee_assignment, which returns None when there is no assignment (phase0 honest validator — Attesting).
Against a local BN this is wasted traffic. Against Obol Charon the endpoint blocks on DutyDB until the cluster has attestation data for a real duty, so an idle slot never returns. The VC then times out and flaps the node:
Against a local BN this is wasted traffic. Against Obol Charon the endpoint blocks on DutyDB until the cluster has attestation data for a real duty, so an idle slot never returns. The VC then times out and marks the beacon node down (it is logged online again shortly after):

t+0.000s  INF  Slot start        node_status=synced delay=~2ms
t+7.5s    WRN  Beacon node down  reason="Timeout exceeded … produceAttestationData(best)"
t+7.5s    WRN  Unable to proceed attestations  duties_count=0  service=attestation_service

duties_count=0 means no attester duty this slot (Slot start also shows attestationIn many minutes away). On a duty slot the same path publishes normally (~50ms after attestation-due).

Approach

In spawnAttestationTasksV2, return immediately when getAttesterDutiesForSlot is empty.

Test plan

• Unit tests in tests/test_validator_client.nim for getAttesterDutiesForSlot: empty map, duty on another slot, duty on the requested slot
• CI unit suite / test_validator_client

AI assistance disclosure

I used Cursor to help inspect the Nimbus VC attestation loop, compare other clients and the honest-validator spec, draft this patch, and prepare this PR text. I reviewed the resulting diff and can explain the change. A human author remains responsible for the submitted code (see Status/Logos contributor guidance on AI-assisted work: allowed for drafting; the opener must review, verify, and be able to explain every change; PRs should not be opened automatically by an agent/bot). I am a node operator rather than a regular Nimbus contributor; corrections from maintainers are welcome.

[tersec]

The following files do not have an up-to-date copyright year:
- tests/test_validator_client.nim

i.e.

nimbus-eth2/tests/test_validator_client.nim

Line 2 in 6a0573f

# Copyright (c) 2018-2025 Status Research & Development GmbH

should read 2018-2026

[github-actions]

Unit Test Results

       12 files  ±0    2 888 suites  ±0   1h 7m 46s ⏱️ + 3m 1s
11 461 tests +1    8 269 ✔️ +1  3 192 💤 ±0  0 ❌ ±0 
57 200 runs  +4  51 373 ✔️ +4  5 827 💤 ±0  0 ❌ ±0 

Results for commit 8baaafa. ± Comparison against base commit 0726147.

♻️ This comment has been updated with latest results.

[b0a7]

The following files do not have an up-to-date copyright year:
- tests/test_validator_client.nim

i.e.

nimbus-eth2/tests/test_validator_client.nim

Line 2 in 6a0573f

# Copyright (c) 2018-2025 Status Research & Development GmbH

should read 2018-2026

Thx, fixed

[tersec]

Against Obol Charon the endpoint blocks on DutyDB until the cluster has attestation data for a real duty, so an idle slot never returns.

This seems questionable.