Query prepared cache is shared across scope and cluster without differentiating

The reason this happens, and only on a fresh cluster, is because Go is using prepared statements. The first time that the query is run at the scope level the SDK is picking up the prepared statement created by the cluster level query. Query itself says that this prepared statement doesn't exist (because it doesn't at the scope level) and returns an error, at which point the SDK reprepares. This only happens on a fresh cluster because after the first test run the prepared statement exists on the server. Running "delete from system:prepareds" between each run of this test reproduces the issue every time.