I like the direction.

What if you run multiple separate deno test calls (for example with different flags), and then generate a single coverage report?

What if you run multiple separate deno test calls (for example with different flags), and then generate a single coverage report?

How about adding some option (like --merge), and if that flag is given, the command automatically look up the results (intermediate files) from the previous runs and merge them.

A merge feature would be nice as it'd allow to combine coverage from different sources too (like when running code in another v8 context, such as browsers when puppeteer)

Also if the summary is being reworked, it may be an opportunity to take a look back at the --threshold in the meantime, I think the previous attempts failed because it was unclear how the total was supposed to be computed

hide internal intermediate json files from the users

I'd like to point out that I'm currently relying on the intermediate files so that I can modify coverage data before generating an lcov file.
Specifically, blob: urls from the json files are replaced with the paths of files that were used to generate the blob: content. (more info here)

But the $coverage_dir/.raw_data suggestion would totally suit my needs since they would still be accessible. I would only need to update the path of these files.

I'd love to see #18147 implemented.