Most games have a debug and a release configuration. Runtime checks (eg. asserts or #ifdef DEBUG) in debug code help discovering errors and they are excluded from the release binary.
So testing is done on the debug configuration to discover all errors. When there are no more unfixed bugs, configuration is switched to release and the game is rebuilt.
How much is needed to be tested in the release version again?
Testing everything all over again would be too much. Not testing anything would be also crazy, as the code changed and it should not change the behavior but in software development people make mistakes...
How would you chose what/how to test again in the release?