Not every company can afford to run crappy code in production. PayPal and banks, for example. Being able to quickly roll back changes while still keeping developer velocity and moving forward is important, and it is very difficult when thousands of changes are going out every day in the same monolith.
But not all forward fixes are quick to write or even locate. Also, monoliths take much longer to go through CI (build + all the tests), and deployment tends to be much slower.
In cases when a rollback is necessay, you'll often find yourself in a cycle where a breakage goes out, the deployment is rolled back, a patch is applied, a new deployment goes out and a new breakage happens, the deployment is rolled back, and so on. I am exaggerating how often this happens, but it does happen when you have thousands of engineers, and it happened often enough at my workplace that developers were unhappy and management pushed for breaking up the monolith.
