All useful software has bugs, and neither the length of the bug list nor the unit test code coverage will tell you whether a test framework meets your needs.
Where I work, groups use either JUnit or TestNG for Java unit testing. We chose them because they are widely used, and so most of the common bugs -- the ones we are mostly likely to encounter -- have already been fixed. Any common bugs that are not fixed are probably well-documented in release notes or in something like StackOverflow.
If I were selecting a unit-testing framework from scratch, I would consider the following:
- Is it open-source? (binary metric)
- Is the documentation adequate?
- Are there lots of examples written by people other than the authors?
- Is it easy to get answers to questions, preferably because someone else already asked them and answered them.
- When people answer questions about the framework, do they tend to sound satisfied with it, or do they sound angry/frustrated?
The question specifically asked for metrics, and most of those points are not phrased quantitatively, but I think you could find ways to quantify most of them.
The Op specifically wants to know whether test/code coverage will tell them anything about test framework reliability. I would be concerned about a framework with no unit tests, but if I were comparing two frameworks that both had unit tests, I would be reluctant to choose one over the other based solely on which one had more test/code coverage.