docs/ch15: mark distributed tracing fully integrated

Every authed API endpoint now produces a nested flame graph
(HTTP → auth → service → SQL). Replaces the in-flight section with the
final span-source matrix and a sample 5-span /api/tasks/ trace. Notes
the visible Hetzner→Neon transatlantic RTT as the perf bottleneck the
flame graph surfaced.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

docs/deployment/15-observability.md

@@ -220,42 +220,54 @@ Cheapest fix path:
 We'd want both eventually. Grafana alerting first because the data is
 already there.
 
-### Distributed tracing — adoption is in flight
+### Distributed tracing — fully integrated
 
-The OTel SDK is **wired** in `cmd/api/main.go` and `cmd/worker/main.go`
+The OTel SDK is wired in `cmd/api/main.go` and `cmd/worker/main.go` and
-and ships traces to Jaeger via `obs.88oakapps.com/v1/traces`. What's
+ships traces to Jaeger via `obs.88oakapps.com/v1/traces`. Every public
-already producing spans:
+service method now takes `ctx context.Context` and routes its SQL through
+`repo.WithContext(ctx)`, which means **every authenticated API endpoint
+produces a fully-nested flame graph** in Jaeger.
 
 | Span source | Status |
 |---|---|
 | `otelecho.Middleware` — span per HTTP request | ✅ live |
+| Auth middleware DB lookups (`m.db.WithContext(ctx)`) | ✅ live |
+| All repos via `repo.WithContext(ctx)` (`otelgorm` plugin) | ✅ live |
 | Manual span around `storage_service.Upload` (B2 PutObject) | ✅ live |
 | Manual span around APNs `Send` / `SendWithCategory` | ✅ live |
 | Manual span around FCM `sendOne` | ✅ live |
 | Asynq middleware — span per task type with retry/payload attrs | ✅ live |
-| `otelgorm` plugin — span per SQL statement | ✅ plugin registered |
 
-What's still in flight: SQL spans appear in a request's trace **only when
+Migrated services (every public method takes ctx):
-the service method took the request's `ctx` and called
+- `AuthService` — login, register, refresh, logout, me, verify-email,
-`repo.WithContext(ctx)`** before issuing queries. Every repository now
+  forgot/reset-password, update-profile
-exposes `WithContext(ctx) *XRepository`, but services need to be
+- `TaskService` — all 25+ task and completion methods
-migrated one method at a time.
+- `ResidenceService` — all 15 methods including share-codes
+- `ContractorService` — all 9 methods
+- `DocumentService` — all 10 methods
+- `NotificationService` — all 12 methods
+- `SubscriptionService` — all 12 methods including Apple/Google IAP
 
-**Migration pattern:** for each service method on the request hot path,
+Sample trace for `GET /api/tasks/`:
-add `ctx context.Context` as the first arg, change the handler call site
+
+```
+GET /api/tasks/                                              (2473ms)
+├── auth: SELECT * FROM user_authtoken WHERE key=...           (1506ms)
+├── auth: SELECT * FROM auth_user WHERE id=7                    (333ms)
+├── service: SELECT id FROM residence_residence WHERE...        (736ms)
+└── service: SELECT * FROM task_task WHERE residence_id IN(...) (226ms)
+```
+
+Each query labeled with `db.statement`. The transatlantic Hetzner→Neon
+RTT (~110ms one-way) shows up clearly — that's the perf bottleneck the
+flame graph makes obvious. See [Chapter 18 — Cost](./18-cost.md) for
+the planned move to a US-region cluster.
+
+**Migration pattern (for any future services or middleware):** add
+`ctx context.Context` as the first arg, change the handler call site
 to pass `c.Request().Context()`, and replace `s.repo.X(...)` with
 `s.repo.WithContext(ctx).X(...)`. Tests pass `context.Background()`.
 
-Already migrated:
-- `TaskService.ListTasks` → `GET /api/tasks/`
-- `TaskService.GetTasksByResidence` → `GET /api/tasks/by-residence/:id/`
-
-Remaining: every other public method on `TaskService`, `ResidenceService`,
-`ContractorService`, `DocumentService`, `AuthService`,
-`NotificationService`, `SubscriptionService`. Mechanical work; can be
-done a method at a time without breaking anything (untouched methods
-just emit untraced SQL like before).
-
 ### No APM (Application Performance Monitoring)
 
 No continuous profiling. We can answer "which endpoint has the highest