diff --git a/src/donkeyq/client.py b/src/donkeyq/client.py
index 1b4293d..73b2287 100644
--- a/src/donkeyq/client.py
+++ b/src/donkeyq/client.py
@@ -23,10 +23,10 @@ async def handle(msg):
 import socket
 from collections.abc import Callable
 from dataclasses import dataclass
-from typing import Any, cast
+from typing import TYPE_CHECKING, Any, cast
 
 import redis.asyncio as aioredis
-from redis.exceptions import ResponseError
+from redis.exceptions import RedisError, ResponseError
 
 from . import keys
 from .connection import make_client
@@ -37,6 +37,9 @@ async def handle(msg):
 from .scripts import register_redrive
 from .serde import Schema, resolve_schema, to_jsonable
 
+if TYPE_CHECKING:
+    from pydantic import BaseModel
+
 logger = logging.getLogger(__name__)
 
 # Stable for this process, unique across processes/hosts — so workers sharing a
@@ -200,7 +203,7 @@ def _register(
         on_dlq: OnDLQ | None = None,
         block_ms: int = 2000,
         count: int = 10,
-        schema: Any = None,
+        schema: type[BaseModel] | Schema | None = None,
         start_from: str = "$",
         dedupe_ttl_s: float | None = None,
     ) -> _Subscription:
@@ -410,8 +413,18 @@ async def _register_in_registry(self, sub: _Subscription) -> None:
         )
         try:
             await self._require_client().sadd(self._registry_key, record)
-        except Exception:
-            logger.warning("failed to write registry entry for %s/%s", sub.stream, sub.group)
+        except RedisError:
+            # The registry is best-effort metadata (cross-stream enumeration for a
+            # future dashboard), not on the delivery path — a Redis hiccup here must
+            # not fail activation. Log with the cause so it's diagnosable; anything
+            # that isn't a RedisError is a real bug and is left to propagate.
+            logger.warning(
+                "failed to write registry entry for %s/%s; dashboard enumeration may "
+                "miss this subscription (delivery is unaffected)",
+                sub.stream,
+                sub.group,
+                exc_info=True,
+            )
 
     # -- publish -----------------------------------------------------------
 
@@ -422,7 +435,7 @@ async def publish(
         *,
         headers: dict[str, Any] | None = None,
         maxlen: int | None = None,
-        schema: Any = None,
+        schema: type[BaseModel] | Schema | None = None,
     ) -> str:
         """Publish ``body`` to ``stream``. Returns the new entry id.
 
diff --git a/src/donkeyq/reclaimer.py b/src/donkeyq/reclaimer.py
index 47d70fb..6daf71e 100644
--- a/src/donkeyq/reclaimer.py
+++ b/src/donkeyq/reclaimer.py
@@ -309,28 +309,37 @@ async def _reclaim_entry(
         )
         if action in ("dlq", "poison_dlq"):
             logger.warning(
-                "message %s routed to DLQ %s (%s, count=%s)",
+                "message %s on %s (group %s) routed to DLQ %s (%s, count=%s)",
                 msg_id,
+                config.stream,
+                config.group,
                 dlq,
                 action,
                 new_count,
             )
             await self._invoke_on_dlq(config, body_for_cb, original_id, new_count)
         elif action == "poison_drop":
-            logger.error("message %s has an unreadable envelope and no DLQ; dropped", msg_id)
+            logger.error(
+                "message %s on %s (group %s) has an unreadable envelope and no DLQ; dropped",
+                msg_id,
+                config.stream,
+                config.group,
+            )
         elif action == "gone":
             logger.warning(
-                "message %s on %s was trimmed from the stream while still pending; "
-                "PEL cleared and the message is LOST — a retry/stream MAXLEN cap is "
-                "set below the in-flight backlog",
+                "message %s on %s (group %s) was trimmed from the stream while still "
+                "pending; PEL cleared and the message is LOST — a retry/stream MAXLEN "
+                "cap is set below the in-flight backlog",
                 msg_id,
                 config.stream,
+                config.group,
             )
         elif action == "retry":
             logger.debug(
-                "reclaimed message %s on %s → retry stream %s (count=%s)",
+                "reclaimed message %s on %s (group %s) → retry stream %s (count=%s)",
                 msg_id,
                 config.stream,
+                config.group,
                 config.retry_stream,
                 new_count,
             )