Need a Clarification on function sequences

When I built my agent using ML-Agents, I didn't use a decision requestor component. Instead I used the in-built RequestDecision() function to get a decision. My question is what is the order of the functions are called.

My thoughts that it was like something like this.(After an Episode Begin)

RequestDesicion() -----> CollectObservations----> OnActionsReceived

But I noticed that the CollectObservations function is being called even I am not requesting for a decision.
So my questions are,

1. What is the exact moment the CollectObservations is called?
2. Does it affect my training if the function is being called even without request a Decision?

I am using RequestDecision() only once during an Episode, but CollectObservation is called twice.

3. Does the collectObservation function is called twice during an episode?


Here is the logs of functions called during a Heuristic mode. (Note that I haven't added a log to the Heuristic function & cannot access the RequestDecision, so the log on that was moment before it was being called)

Most importantly I want to know is how does this affect my training on the Agent?