How does retrying failed flow runs intersect with new flow run code?

In our Prefect 2 setup we have set PREFECT_RESULTS_PERSIST_BY_DEFAULT=true to persist results. If the code defining the flow (stored in a GCP bucket) has been updated since the failed run, when we click Retry on a failed run, will the retried flow run use the old version of the code for the flow or the new one?

Curious if anyone knows about this? I guess I need to figure it out myself