Spotted using the logging from #7260 The tailnet coordinator sometimes drops a `Node` update, causing connections to agents to fail to be established. [TestAgent_EnvironmentVariables.txt](https://github.com/coder/coder/files/11332185/TestAgent_EnvironmentVariables.txt) See above logs for detailed sequence. It's a classic concurrency bug: 1. Grab the latest data 2. Subscribe to updates If an update comes in between 1 and 2, we miss it.