If you send the pool upgrade transaction and set the version number to something higher than what the pool has installed, but to a version that does not exist the node get stuck in an upgrade loop.
So if your node version is 0.3.24 and the newest version if 0.3.25 you send the command with "version=0.3.50". Since 0.3.50 does not exist the nodes will loop trying to upgrade.
You can see this by watching the CLI when it is connected to the pool it will start to repeat disconnects and connects.
The logs are large and it is easy to reproduce so I am not including them.