chore(earnings): enable the earnings indexer cron (#978)

[biwasxyz]

Flips the earnings indexer on by setting EARNINGS_INDEX_ENABLED=true (top-level + env.production, omitted from env.preview so preview never indexes against shared prod D1/Hiro). Mirrors the TENERO_REFRESH_ENABLED pattern.

Verified working before enabling

Ran a forced admin sweep against prod (POST /api/admin/scheduler?action=refresh&task=earnings) after applying migrations 020–022:

• 25 agents, 31 Hiro calls (≈1.2/agent — join-date floor + bounded backfill working)
• 116 earning rows, ~$1,666 verified (sBTC + STX), all priced
• 6 self_funded exclusions (anti-gaming fired), 186 unclassified correctly excluded
• alreadyKnown: 92 on re-run → idempotency confirmed

What happens on merge/deploy

The 30-min cron sweep starts: 25 agents/tick → ~20h for the first full backfill of the ~1k roster, then cheap incremental cycles (only-new-txs per agent, often zero work). Monitor via GET /api/admin/scheduler → .status.lastEarningsResult.

Flip back off any time by reverting this (or wrangler secret put). Part of #978.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	landing-page	3a889cf	Commit Preview URL Branch Preview URL	Jun 08 2026, 09:22 AM

[arc0btc]

Enables the earnings indexer cron by setting EARNINGS_INDEX_ENABLED=true in top-level and env.production — intentionally omitted from env.preview to protect shared prod D1/Hiro quota.

What looks good:

• The pre-merge manual sweep is exactly the right gate here: 25 agents, 31 Hiro calls (≈1.2/agent confirms the join-date floor is doing its job), idempotency verified via alreadyKnown: 92 on re-run. That's a complete pre-flight.
• Mirrors the established TENERO_REFRESH_ENABLED pattern — consistent with the repo's flag convention.
• Preview env exclusion is correct: preview builds should never index against shared production D1/Hiro quota, and omission (rather than false) is the right approach since a missing env var is easier to audit than a stale false.
• The circuit-breaker from PR #958 is already in place — Hiro call storms during the ~20h backfill window are covered.

[nit] The inline comment on line 54 reads Ships off; flip to "true" to start indexing but the value is already "true". That comment was accurate when the flag was being introduced as off-by-default; now that this PR is enabling it, the comment is misleading. Consider updating to something like Enabled: 30-min earnings indexer cron sweep (issue #978). or just removing it since the PR description and issue #978 already document the history.

Code quality notes:

• The second block (line 158, env.production) correctly mirrors the top-level vars — no gaps there.

Operational note: The 30-min cron cadence against ~1k roster with bounded incremental cycles after backfill is a reasonable pattern. The backfill math (~20h at 25 agents/tick) is realistic and won't spike Hiro quota in a way the circuit-breaker can't absorb.