If computers could read thoughts and 100% reliably determine when the user is done with their current task and would like to switch back to the other application, then yes, it would definitely make sense to implement the functionality you're describing.

In the real world, where it's impossible for the OS to distinguish between the "You stopped checking in the first 0.5 sec while the app took 1 sec to start" scenario and the "you're still busy with the current task and don't want to switch to the other app yet" scenario, the safe choice is always to not steal the focus because the cost of switching too early is significantly higher than the cost of not switching.