I'm following a best practice of inserting a smart wait for an HTML mode element prior to reading/navigating it. This is very reliable, and my attributes are as refined as possible.
After successfully clearing the wait stage, I can rarely receive an error in the Read or Navigate during runtime, such as:
"Failed To perform Step 1 In Read Stage '' on page '' - Unable to match any windows with the query terms"
Suppose my process loops through this stage for 1000 records, my bot can handle this successfully during iterations 1-600, encounter this error on iteration 601, and be successful from 602-1000. The strange thing is that I can process record 601 in the exact same way during a subsequent run, and it will be successful without hitting this error. Nothing's really too complex about the process, only one window of the target application is ever open at a time, and the element is fairly unique.
I can put in retry logic for when it fails, and this generally resolves the problem. But my client would like to know why this happens for the purpose of having to deploy subsequent releases to production because of this issue. Can I get an explanation for why this element would pass the wait stage, but that same element (although rarely) fails the navigate or read stage that comes immediately after?