diff --git a/reg-tests/mcli/mcli_master_socket_leak.vtc b/reg-tests/mcli/mcli_master_socket_leak.vtc
deleted file mode 100644
index cf30ea6fe..000000000
--- a/reg-tests/mcli/mcli_master_socket_leak.vtc
+++ /dev/null
@@ -1,66 +0,0 @@
-varnishtest "Bug fix: master CLI connection slots freed on client disconnect"
-
-# Regression test for a master CLI socket connection leak in master-worker mode.
-#
-# mworker_proxy has a fixed maxconn of 10.  When clients connect to the master
-# socket, send a command that is forwarded to a busy worker, and then close
-# the connection due to a client-side receive timeout, haproxy must free each
-# slot as the client disconnects.
-#
-# Without the fix: slots remain occupied after the client disconnects, so the
-# master CLI becomes unreachable once all 10 slots are filled.
-# With the fix: slots are freed on client disconnect and the master CLI keeps
-# accepting new connections.
-#
-# The worker is made unresponsive using "debug dev delay" (expert-mode) with
-# nbthread 1 so a single delay blocks the entire worker CLI.  This avoids
-# using SIGSTOP/SIGCONT which can be unreliable in CI environments.
-
-#REGTEST_TYPE=bug
-
-feature cmd "command -v socat"
-feature cmd "command -v timeout"
-feature ignore_unknown_macro
-
-server s1 {
-} -start
-
-haproxy h1 -W -S -conf {
-    global
-        nbthread 1
-
-    defaults
-        mode http
-        timeout connect "${HAPROXY_TEST_TIMEOUT-5s}"
-        timeout client  "${HAPROXY_TEST_TIMEOUT-5s}"
-        timeout server  "${HAPROXY_TEST_TIMEOUT-5s}"
-
-    frontend fe
-        bind "fd@${fe}"
-        default_backend be
-
-    backend be
-        server s1 ${s1_addr}:${s1_port}
-} -start
-
-# Fill all 10 master CLI slots (mworker_proxy->maxconn is hardcoded to 10).
-# Each socat sends "expert-mode on" followed by "@1 debug dev delay 10000"
-# which blocks the single worker thread for 10 s.  After 2 s the timeout(1)
-# wrapper kills socat, simulating a client-side receive timeout.  "wait"
-# ensures all background processes have exited before proceeding.
-shell {
-    for i in $(seq 1 10); do
-        (printf "expert-mode on\n@1 debug dev delay 10000\n" \
-         | timeout 2 socat TCP:${h1_mcli_addr}:${h1_mcli_port} - 2>/dev/null) &
-    done
-    wait
-}
-
-# This is the key assertion: after all 10 clients have disconnected, a new
-# connection to the master CLI must succeed.  With the bug all 10 slots are
-# still marked occupied and this connect is refused or times out.
-haproxy h1 -mcli {
-    send "show version"
-    expect ~ "3."
-}
-
diff --git a/src/cli.c b/src/cli.c
index 55cd8dab8..8e96b4cd3 100644
--- a/src/cli.c
+++ b/src/cli.c
@@ -3907,9 +3907,8 @@ int mworker_cli_create_master_proxy(char **errmsg)
 	mworker_proxy->mode = PR_MODE_CLI;
 	/* default to 10 concurrent connections */
 	mworker_proxy->maxconn = 10;
-	mworker_proxy->timeout.client = 0; /* no timeout */
-	mworker_proxy->timeout.serverfin = MS_TO_TICKS(1000); /* 1s timeout in case worker is not responding on shutdown */
-
+	/* no timeout */
+	mworker_proxy->timeout.client = 0;
 	mworker_proxy->conf.file = strdup("MASTER");
 	mworker_proxy->conf.line = 0;
 	mworker_proxy->accept = frontend_accept;