While this is posed as very solutions-oriented, this is to consider removal of integration tests from coverage.

These are very slow and expense to run with coverage instrumentation.

These are also expensive to maintain. The infra currently is not able to use bazel yet for resmoke tasks (a separate effort captured in later phases of DEVPROD-15696), so we have to compile with extra flags, manage the .gcno/.gcda files carefully, use gcovr to process the files into json, and then post that data to Coveralls.