diff --git a/lisa/tools/chrony.py b/lisa/tools/chrony.py
index 34cdf01d30..fe62147451 100644
--- a/lisa/tools/chrony.py
+++ b/lisa/tools/chrony.py
@@ -26,9 +26,7 @@
 
 class Chrony(Tool):
     # Leap status     : Normal
-    __leap_status_pattern = re.compile(
-        r"([\w\W]*?)Leap status.*:.*Normal$", re.MULTILINE
-    )
+    __leap_status_pattern = re.compile(r"Leap status\s*:\s*Normal")
     __no_server_set = "Number of sources = 0"
     __service_not_ready = "503 No such source"
 
@@ -69,7 +67,7 @@ def restart(self) -> None:
     def check_tracking(self) -> None:
         cmd_result = self.run("tracking", force_run=True, sudo=True)
         cmd_result.assert_exit_code()
-        if not self.__leap_status_pattern.match(cmd_result.stdout):
+        if not self.__leap_status_pattern.search(cmd_result.stdout):
             raise LisaException(
                 f"Leap status of {self.command} tracking is not expected,"
                 " please check the service status of chrony."
diff --git a/lisa/tools/dhclient.py b/lisa/tools/dhclient.py
index a4c1978b3a..1c49f6c40a 100644
--- a/lisa/tools/dhclient.py
+++ b/lisa/tools/dhclient.py
@@ -5,7 +5,7 @@
 
 from lisa.base_tools import Cat
 from lisa.executable import Tool
-from lisa.operating_system import Debian, Fedora, Redhat, Suse
+from lisa.operating_system import CBLMariner, Debian, Fedora, Redhat, Suse
 from lisa.util import LisaException, UnsupportedDistroException, find_group_in_lines
 
 from .ls import Ls
@@ -71,10 +71,32 @@ def get_timeout(self) -> int:
             if group and not group["default"]:
                 value = int(group["number"])
                 is_default_value = False
-        elif isinstance(self.node.os, Fedora):
-            # the default value in fedora is 300
+        elif isinstance(self.node.os, (Fedora, CBLMariner)):
+            # Fedora and AZL4+ use NetworkManager for DHCP; default timeout is 300.
+            # CBLMariner < 4 is not supported here and raises below.
+            if isinstance(self.node.os, CBLMariner) and (
+                self.node.os.information.version.major < 4
+            ):
+                raise UnsupportedDistroException(
+                    os=self.node.os,
+                    message=(
+                        "CBLMariner < 4 (AZL2/3) uses dhcpcd rather than "
+                        "NetworkManager, so this NetworkManager-based timeout "
+                        "lookup does not apply. To add support, read the timeout "
+                        "from the dhcpcd configuration (e.g. /etc/dhcpcd.conf) "
+                        "instead."
+                    ),
+                )
             value = 300
-            result = self.node.execute("NetworkManager --print-config", sudo=True)
+            result = self.node.execute(
+                "NetworkManager --print-config",
+                sudo=True,
+                expected_exit_code=0,
+                expected_exit_code_failure_message=(
+                    "Failed to read DHCP timeout from NetworkManager. "
+                    "Ensure NetworkManager is installed and accessible."
+                ),
+            )
             group = find_group_in_lines(result.stdout, self._fedora_pattern)
             if group and value != int(group["number"]):
                 value = int(group["number"])
diff --git a/lisa/tools/kdump.py b/lisa/tools/kdump.py
index 8aa0c1891b..af0c510309 100644
--- a/lisa/tools/kdump.py
+++ b/lisa/tools/kdump.py
@@ -226,6 +226,8 @@ def create(cls, node: "Node", *args: Any, **kwargs: Any) -> Tool:
         elif isinstance(node.os, Suse):
             return KdumpSuse(node)
         elif isinstance(node.os, CBLMariner):
+            if node.os.information.version.major >= 4:
+                return KdumpFedora(node)
             return KdumpCBLMariner(node)
         elif type(node.os) is Fedora:
             return KdumpFedora(node)
@@ -585,7 +587,7 @@ def command(self) -> str:
         return "kdumpctl"
 
     def _install(self) -> bool:
-        assert isinstance(self.node.os, Fedora)
+        assert isinstance(self.node.os, (Fedora, CBLMariner))
         self.node.os.install_packages("kdump-utils")
         return self._check_exists()
 
diff --git a/lisa/tools/kill.py b/lisa/tools/kill.py
index e26cbbb470..15fe0dc355 100644
--- a/lisa/tools/kill.py
+++ b/lisa/tools/kill.py
@@ -6,6 +6,7 @@
 from lisa.util import LisaException
 from lisa.util.constants import SIGKILL
 
+from .pgrep import Pgrep
 from .pidof import Pidof
 
 
@@ -21,17 +22,19 @@ def can_install(self) -> bool:
     def by_name(
         self, process_name: str, signum: int = SIGKILL, ignore_not_exist: bool = False
     ) -> None:
-        # attempt kill by name first
-        kill_by_name = self.run(
-            f"-s {signum} {process_name}", sudo=True, shell=True, force_run=True
-        )
-        if kill_by_name.exit_code == 0:
+        # Prefer pidof for exact-name matching to avoid regex side-effects,
+        # then fall back to pgrep for processes started via sudo/shell wrappers.
+        pids = self.node.tools[Pidof].get_pids(process_name, sudo=True)
+        if pids:
+            for pid in pids:
+                self.by_pid(pid, signum, ignore_not_exist)
             return
 
-        # fallback to kill by pid if first attempt fails for some reason
-        pids = self.node.tools[Pidof].get_pids(process_name, sudo=True)
-        for pid in pids:
-            self.by_pid(pid, signum, ignore_not_exist)
+        # Fallback: try pgrep in case pidof missed it (e.g. wrapper processes)
+        processes = self.node.tools[Pgrep].get_processes(process_name)
+        if processes:
+            for proc in processes:
+                self.by_pid(proc.id, signum, ignore_not_exist)
         else:
             self._log.debug(
                 f"Kill for {process_name} did not find any processes to kill."
diff --git a/lisa/tools/modprobe.py b/lisa/tools/modprobe.py
index 7b8a0b2cb1..3ebcca169e 100644
--- a/lisa/tools/modprobe.py
+++ b/lisa/tools/modprobe.py
@@ -242,6 +242,11 @@ def reload(
                     "\nTrying to reconnect to the remote node in 2 sec..."
                 )
                 time.sleep(2)
+                # Reset the stale SSH transport so the next retry establishes
+                # a fresh connection. Without this, the retry loop keeps
+                # reusing the broken session and every attempt fails with
+                # "cannot connect to TCP port 22".
+                self.node.close()
 
         self._log.debug(
             f"Time taken to reload {mod_name}: {timer.elapsed(False)} seconds"