From d85161f1419da418fa3f3127aa0b275237374104 Mon Sep 17 00:00:00 2001
From: TaprootFreak <142087526+TaprootFreak@users.noreply.github.com>
Date: Wed, 20 May 2026 14:10:25 +0200
Subject: [PATCH] Active BTC heartbeat to stop chain-outage false positives
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

`checkBtcStaleness` alerts when `btcLastSuccessMs` is older than 1h. The
field was only advanced by `getBtcPriceInUsd`, which is only reached
when the monitoring cycle prices a WCBTC position. When the chain stops
producing blocks (Citrea outage 2026-05-19 07:40–09:30 UTC, ~115 min),
the cycle has no new positions to price → the on-demand path never
runs → the hourly probe at 09:00 UTC fires a CRITICAL Telegram alert
for a problem entirely off the BTC-spot path.

This adds an active heartbeat that fetches the BTC spot every 5 minutes
regardless of chain activity. The watchdog now only fires when CoinGecko
or the pricing-proxy is actually unreachable.

`getBtcPriceInUsd` and the new heartbeat share `fetchAndCacheBtcPrice`,
which does the upstream call, populates the cache, advances
`btcLastSuccessMs`, and clears the alert dedup timer on success.

Five-minute cadence matches the existing `MonitoringService` cron
interval, well inside `STALENESS_ALERT_THRESHOLD_MS = 60 min`. Each
heartbeat hits the in-cluster pricing-proxy with its 60 s cache so the
upstream cost is at most ~12 CoinGecko-Pro calls/hour per monitor
container (still subject to the existing PRICE_CACHE_TTL_MS).
---
 src/monitoringV2/price.service.ts | 28 +++++++++++++++++++++++++++-
 1 file changed, 27 insertions(+), 1 deletion(-)

diff --git a/src/monitoringV2/price.service.ts b/src/monitoringV2/price.service.ts
index b4963a0..7b68aca 100644
--- a/src/monitoringV2/price.service.ts
+++ b/src/monitoringV2/price.service.ts
@@ -135,6 +135,17 @@ export class PriceService {
 			return cached.value;
 		}
 
+		const fresh = await this.fetchAndCacheBtcPrice();
+		if (fresh !== null) return fresh;
+		return cached?.value ?? null;
+	}
+
+	/**
+	 * Unconditional fetch + cache + bookkeeping. Used both by the on-demand
+	 * `getBtcPriceInUsd` (after cache miss) and the active heartbeat below.
+	 * Returns null on upstream failure; never throws.
+	 */
+	private async fetchAndCacheBtcPrice(): Promise<string | null> {
 		try {
 			const { baseUrl, headers } = this.resolveCoingeckoEndpoint();
 			const response = await axios.get(`${baseUrl}/api/v3/simple/price?ids=bitcoin&vs_currencies=usd`, {
@@ -151,10 +162,25 @@ export class PriceService {
 			return price;
 		} catch (error) {
 			this.logger.error(`Failed to fetch BTC price: ${error.message}`);
-			return cached?.value ?? null;
+			return null;
 		}
 	}
 
+	/**
+	 * Active BTC heartbeat — fires every 5 minutes regardless of chain
+	 * activity. Without this, `btcLastSuccessMs` only advances when the
+	 * monitoring cycle pulls a WCBTC position price; if the chain stops
+	 * producing blocks (Citrea outage), the on-demand path never runs and
+	 * the hourly staleness watchdog below fires a false positive for a
+	 * problem that is entirely off the BTC-spot path. The heartbeat
+	 * decouples the watchdog from the block tick: it now only alerts when
+	 * CoinGecko/the pricing-proxy is actually unreachable.
+	 */
+	@Cron(CronExpression.EVERY_5_MINUTES)
+	async refreshBtcPriceHeartbeat(): Promise<void> {
+		await this.fetchAndCacheBtcPrice();
+	}
+
 	/**
 	 * Hourly probe: when the last successful BTC fetch is older than
 	 * STALENESS_ALERT_THRESHOLD_MS, the suspicious-liq-price trigger for WCBTC