Improve consequest count mismatch error behaviour

[therealryan]

When the model expects a different number of downstream requests than are encountered, we throw an error.
The way that the error seems to stop the population of other captured data into the report.

We should find a way to defer raising the error until flow processing is complete.

[therealryan]

Related issue: it looks like, while we get an error is we capture an unexpected number of request, no error is raised if the expected number of requests is zero.