diff --git a/administration_commands.py b/administration_commands.py
index c964d48..cb543f5 100644
--- a/administration_commands.py
+++ b/administration_commands.py
@@ -171,6 +171,17 @@ class AdministrationModule(commands.Cog):
         """
         # logger.info("Heartbeat of cleanup proc")
         channel = self.bot.get_channel(1062047367337095268)
+        if channel is None:
+            # get_channel() returns None before the gateway cache is populated
+            # (the before_loop below normally prevents that) or when the bot
+            # cannot see the channel at all (wrong id / not in the guild /
+            # missing permission). Skip this tick instead of crashing - an
+            # unhandled exception here stops the whole loop, killing log
+            # rollover and the spontaneous messages with it.
+            self.logger.warning(
+                "check_self: channel 1062047367337095268 unavailable - skipping tick"
+            )
+            return
         messages = [message async for message in channel.history(limit=1)]
         for mess in messages:
             channel = mess.channel
@@ -254,6 +265,13 @@ class AdministrationModule(commands.Cog):
                 self.logger.info(message)
                 await channel.send(message)
 
+    @check_self.before_loop
+    async def before_check_self(self):
+        # Don't fire the first tick until the gateway is READY and the channel
+        # cache is populated - get_channel() returns None before that, which is
+        # exactly what used to crash check_self at startup.
+        await self.bot.wait_until_ready()
+
 
 async def setup(bot):
     logger = logging.getLogger("discord")