Secrets Manager rotation failure retry details

In the secrets manager documentation it states:

If any rotation step fails, Secrets Manager retries the entire rotation process multiple times.

I've also seen conflicting reports that it will retry the current step of rotation that it was on.

Is it possible to more clearly document how Secrets Manager retries rotation failures, with specific details? When writing a custom rotation lambda to manage resources that don't fall into one of the generic templates AWS provides, it is necessary to understand how lambda invocation failures are handled by the Secrets Manager service. Without this information, it is impossible to write robust rotation logic that can handle transitive failures of the rotation lambda.