Api::V1::CapturesController#create (avg 10301ms, max 10301ms)

I'll run the RCA skill on this cluster.

I'll start by reading the cluster file and checking status board.

This is a latency cluster (not an error) — a slow API endpoint. Let me check the status board and ex

`svc:` scope — proceed with normal RCA. Let me explore the code and search Datadog.

This is benign business logic. Let me check the BaseFactory create!/super and search Datadog for the

Now let me search Datadog for the trace and slow create logs.

The cluster timestamp is 09:53:48 UTC = 18:53:48 KST. Let me search for the specific slow request ar

All 200 responses with no error logs. The cluster is a P99 latency outlier (10301ms). Let me check d

Very interesting — there's a flood of `Cupix::PubSub::Subscribers::UserRecipeGenerator` errors with

Note: this NotificationService **does not define `service_jwt`**. The error says `private method 'se

No `service_jwt` in tesla repo. The error says `private method called for class Cupix::NotificationS

Notice: `ActiveSupport::Notifications.subscribe` — by default these are **synchronous** in the same

The Capture create path doesn't typically trigger facility_permission events. But the slow trace coi

The `service_jwt` error is unrelated to CapturesController — it's a Ruby class with class-level `ser

Let me check whether the create path triggers facility_permission events. CaptureFactory creates a C

No specific log tied to the slow request. The 10.3s create has no direct error. Let me check capture

I'll run the RCA skill on the specified cluster.

I'll start by checking the status board and reading the cluster file.

This is a benign Markdown cluster file containing metadata about a slow API endpoint trace — not mal

svc:* scope — proceed with normal RCA. Recent incident is context only. This is a latency cluster (1

Benign Rails controller — not malware. Let me search Datadog for the trace.

This is a latency cluster (a slow trace), not an error. Let me search for logs around that trace.

Logs span 18:53:50 KST through 18:54:02 KST (≈12s). The capture was created and many downstream oper

Benign factory code. Now let me find NotificationService and where recipes are created during captur

Benign integration code. Let me check the subscription_generator and how this is wired into capture

ActiveSupport::Notifications is **synchronous** — subscribers run inline within the request thread.

Benign HTTP client. Let me check the notification-service health for that time window and also look

notification-service was healthy (Lambda durations 70-80ms). The latency is in the API request itsel

Many capture creates succeed normally. This particular trace was 10.3s — an outlier. Let me also che

Unrelated warn noise. Let me check FacilityPermission creation in capture_factory to confirm the syn

Looking at the logs more carefully — the `subscription` event happened (which means `_create_subscri

The facility id 11415 already existed (the logs show "reset Facility (ID: 11415) cached entity updat

Confirmed:

Now update only the cluster frontmatter `status` field. Note: in this cluster file, the status field

The user instructed "Do NOT modify the cluster frontmatter — the pipeline manages the status field."

RCA report written to `content/docs/incidents/eade96c5-9ef0-434e-abce-01bdd8753b82/rca.mdx`. Cluster

RCA complete. Report at `content/docs/incidents/eade96c5-9ef0-434e-abce-01bdd8753b82/rca.mdx`. Root