Buggy output tests currently take into account the output for all tasks, whereas they should only consider the output for the task they're defined in.

[seanlip]

Suppose a question has two tasks, task1 and task2. If a student's code matches a reference buggy-output code for all test suites in task1, but not all test suites in task2, then the code is deemed not to match the buggy-output code in general, and will not get the targeted feedback in task1.

This is incorrect behaviour. The "matches buggy-output" criterion should only consider the code from the given task.

(This bug was discovered as part of the audit of the ConversationManagerService integration tests in #632.)