diff --git a/bioengine/_version.py b/bioengine/_version.py
index 83f80f7..af6c015 100644
--- a/bioengine/_version.py
+++ b/bioengine/_version.py
@@ -13,4 +13,4 @@
 Must stay in lock-step with ``pyproject.toml``'s ``version`` field.
 The ``version-check.yml`` CI workflow enforces the match.
 """
-__version__ = "0.11.25"
+__version__ = "0.11.26"
diff --git a/bioengine/apps/proxy_deployment.py b/bioengine/apps/proxy_deployment.py
index e97b659..705719b 100644
--- a/bioengine/apps/proxy_deployment.py
+++ b/bioengine/apps/proxy_deployment.py
@@ -428,6 +428,26 @@ async def deployment_function(*args, context: Dict[str, Any], **kwargs) -> Any:
             # Semaphore bounds concurrency for both WebSocket and WebRTC entry paths —
             # they share this single wrapper, so a long-running call from either
             # transport consumes one slot until it returns.
+            #
+            # Fail fast instead of queueing unboundedly when at capacity: a
+            # blocked caller still holds its decoded payload (e.g. an inference
+            # image) in this actor's memory while it waits, so an open-ended
+            # backlog grows proxy memory without limit. Because we reject rather
+            # than await, the semaphore never accumulates waiters — so
+            # ``.locked()`` means "no free slot", and there is no ``await``
+            # between this check and the acquire below, making the decision
+            # race-free.
+            if self.service_semaphore.locked():
+                logger.warning(
+                    f"⚠️ '{self.application_id}' at capacity "
+                    f"({self.max_ongoing_requests} concurrent requests); "
+                    f"rejecting '{method_name}'."
+                )
+                raise RuntimeError(
+                    f"Application '{self.application_id}' has reached its maximum "
+                    f"of {self.max_ongoing_requests} concurrent requests. Please "
+                    f"try again in a moment."
+                )
             async with self.service_semaphore:
                 try:
                     await self._check_permissions(context, method_name=method_name)
diff --git a/pyproject.toml b/pyproject.toml
index 12eb65a..80fcbbd 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "bioengine"
-version = "0.11.25"
+version = "0.11.26"
 description = "BioEngine — CLI and SDK for deploying and calling AI model services on BioEngine workers"
 requires-python = ">=3.11"
 authors = [