Reminds me of the engineer who decided to test some circuit resilience during working hours.
We had a process to test resilience on customer sites at least once a year. Resilience testing followed our change process. We needed user approval with a scheduled time window and field engineers on call just in case we somehow managed to completely disconnect the site. But somebody decided that this didn't matter on the grounds that it was going to work.
On this occasion there has been a configuration change and out of the now legendary "abundance of caution" the customer decided they would like a resilience test carrying out in the site in question outside the normal testing schedule. The correct process was for the engineer to raise an RFC for failover testing out of hours. The engineer who had carried out the change decided that they would simply test during working hours.
Now even if the test was successful it wouldn't have gone down well with the customer as the failover was implemented using HSRP. With HSRP there's always going to be a brief interruption to service as everything fails over to the standby connection. But our overconfident change engineer didn't consider that for a moment.
The test was not successful. Or normal test procedure for testing is to drop the primary connection where it hits our core by shutting down the interface. For some reason on this occasion the change engineer decided to log onto the primary router and shut down the WAN interface. Because of some config error there service didn't fail over to the standby, but now our hero couldn't get into the primary to bring the WAN interface back up. No problem, just jump onto the standby and hop across to the primary. Except that the config error that meant that the standby wasn't accessible. The connection was up. You could ping the router. But that was all. It wasn't passing any traffic.
Our hero now went into panic mode. Their first instinct was to try to get a field engineer to site. While they were trying to do this however there customer called in and service desk went thorough the usual checks for a down site one of which is to reboot the router. Luckily as there config hadn't been saved with the WAN interface shut down this restored service.