Every configurations must has some unusual things, like there is always a plan B in our life.
We defined a standard version, and call those deviates from this normal route are exceptions.
System errors, unexpected businesse senarios, data error...anyone of these will need to human interventoin.

If your process is not suiatable for RPA, a developer will do a lot fragile scripts to achieve those tricky requirment, however, it is easy to start celebrating once your robot successfully
automates a process for the first time. However, the real work begins in the testing phase to ensure the robot handles exceptions properly.
How to manage an exception during the automation processing? what's the debugging approach?

To minimize the risk for RPA, not only matter of developer's skill, the most important is stakeholder and product owner should understand and do not interfere the design centric apporoach.

Design Centric principles

1. Not go to details first
   Starting with designing high-level workflow for processes to be automated in a language that is understood by the product owner, and not focusing straight into bottom details developement by recording an isolated bot.
2. Common objects
   Setting universal objects by functions that represents demands once implemented in code provides a library of automated reusable modules or business objects that can be re-used across multiple automated process bots. This reduces the level of maintenance needed once automated process are released into production.
3. The use of environment variables and consistant locator
   If the automated process will be changed in future (like URL,filename, email id, date... etc) and developer don't want to touch the code for such changes, try to exposed common factors or specific data to environment.