Question on using package:coverage granularity for tooling

[Deepansh Gupta]

Hi everyone,

While reviewing `package:coverage` in the context of the GSoC idea around
coverage-guided tooling, I had a small design question.

For external tools consuming coverage data, is the intended stable interface
the processed JSON summaries, or is it acceptable for experimental tools to
work closer to raw hit maps / line execution data?

I’m trying to understand where the maintainers expect tooling experiments
to sit abstraction-wise, especially when making heuristic decisions about
which uncovered code paths are worth testing.

Thanks in advance, and apologies if this is already documented somewhere.