docs(architecture): record why single-binary Tempo has its ingester health check off (refs #156)

`verify-tracing` flaked on run 722: `FAIL — no single trace spanned ['bff',
'projection-api']`, green on a plain re-run of the same commit. Not a broken trace
chain — Tempo could not ingest:

    removing distributor_pool failing healthcheck addr=127.0.0.1:9095
      reason="rpc error: code = DeadlineExceeded"
    pusher failed to consume trace data  err="context canceled"   (x18)

Root cause is the mechanism of the data loss, not whatever caused the stall. Tempo
runs single-binary, so distributor and ingester are the *same process* and the
distributor's ingester pool holds exactly one, in-process, member. dskit still
health-checks it over loopback gRPC with a 1s deadline; on the shared runner a
transient stall blows that, the only ingester is dropped from the pool, and every
push fails until the next 15s check interval — spans silently lost. With one
in-process ingester the check can never route around a failure, so it can only ever
discard data.

Fix: `ingester_client.pool_config.healthcheckenabled: false`. This addresses both
candidate triggers (GC pressure near `mem_limit`, CPU contention) at the point where
they turn into lost data, so `mem_limit: 400m` stays untouched — raising it would
risk reintroducing the verify-e2e OOM of #144 on a memory-tight runner.

Also print `tempo_distributor_ingester_clients` on the check's failure path: a
recurrence then names Tempo-dropped-spans instead of costing another container-log
dive, since from the check's side that is indistinguishable from missing
instrumentation.

Verified against the built image: config parses (`-config.verify`), the effective
`/status/config` reports `healthcheckenabled: false`, and the diagnostic reads the
metric off a live Tempo.

docs/architecture/adr-0023-observability-stack.md

@@ -67,6 +67,14 @@ itself, so no in-image healthcheck tool is required.
 - Three more images built each CI run (kept small; not on the health-gate list).
 - Storage is ephemeral container fs — a demo backplane, not a retention target.
   Object storage for Tempo / remote-write for Prometheus is a later concern.
+- Tempo runs **single-binary**, so its distributor and ingester are one process and
+  some of its distributed-mode machinery is not just redundant but harmful. Its
+  ingester-pool health check is disabled (`ingester_client.pool_config`) because with
+  a single in-process ingester the check can never route around a failure — a 1s
+  loopback-gRPC deadline missed under CI load only evicted the one ingester and made
+  Tempo drop spans, which is how `verify-tracing` flaked (#156). Expect the same
+  shape from other distributed-mode knobs if we tune them; the fix is to switch to
+  real multi-ingester Tempo, not to re-enable them here.
 
 ## Coupling rules touched (CLAUDE.md §8)
 

infra/observability/tempo/tempo.yaml

@@ -25,3 +25,15 @@ storage:
       path: /var/tempo/blocks
     wal:
       path: /var/tempo/wal
+
+# #156: don't let the distributor evict its own ingester. Tempo runs single-binary here, so the
+# distributor and the ingester are the same process and the "pool" holds exactly one, in-process,
+# member. dskit still health-checks it over loopback gRPC with a 1s deadline (checkinterval 15s);
+# on the shared CI runner a transient stall blows that deadline, the only ingester is dropped from
+# the pool ("removing distributor_pool failing healthcheck"), and every push then fails ("pusher
+# failed to consume trace data", err="context canceled") until the next check — silently losing
+# spans, which is how verify-tracing flaked. With one in-process ingester the check can never route
+# around a failure, so it can only ever discard data. Turn it off.
+ingester_client:
+  pool_config:
+    healthcheckenabled: false

infra/tracing-check.py

@@ -59,6 +59,20 @@ def services_in_trace(trace_id):
     return names
 
 
+def tempo_ingest_state():
+    """#156: distinguish a broken trace chain from Tempo dropping spans. `ingester_clients` is 0
+    when the distributor has evicted its (single, in-process) ingester over a failed loopback
+    health check — pushes fail and spans are lost, which looks identical to missing instrumentation
+    from here. Diagnostics only; never fails the check."""
+    try:
+        for line in _get(f"{TEMPO}/metrics").decode().splitlines():
+            if line.startswith("tempo_distributor_ingester_clients "):
+                return f"tempo {line.strip()} (0 = no ingester in the pool — evicted, so pushes\n       are failing and spans are being dropped; see #156)"
+    except Exception as e:
+        return f"tempo /metrics unreadable: {e}"
+    return "tempo_distributor_ingester_clients not reported"
+
+
 def main():
     deadline = time.time() + TIMEOUT
     generate_traffic()
@@ -74,6 +88,7 @@ def main():
         generate_traffic()
     print(f"FAIL — no single trace spanned {sorted(WANT)}; services seen: {sorted(seen)}",
           file=sys.stderr)
+    print(f"       {tempo_ingest_state()}", file=sys.stderr)
     return 1