`scala-cli --power config repositories.default` doesn't override global default repo since `1.15.0`

[geny200]

Version(s)
1.15.0 (for 1.14.0 and earlier - OK)

Describe the bug
Hi. We get a problem in our infrastructure: we use private artifactory (as proxy for maven), and on gitlab runner we restrict access to the public Internet. So starting from version 1.15.0, for some reason, the repositories defined via config repositories.default and env COURSIER_REPOSITORIES stopped working - by default, when downloading "Downloading compilation server 2.1.0" (log after commad run), first attempt is "central.sonatype.com ", then "repo.scala-lang.org " and only after all those attempts fail does it try our artifactory.

To Reproduce
We use virtuslab/scala-cli:1.15.0 as base image for gitlab runner, so I'll show how to reproduce at that docker image. <repo> - url for our artifactory

docker run -it  --entrypoint /bin/bash --env COURSIER_REPOSITORIES=<repo> --rm virtuslab/scala-cli:1.15.0
echo -e "//> using scala 3.8.4\n\n@main def foo(): Unit = {}" > test.scala
scala-cli --power config repositories.default <repo>
scala-cli run test.scala

so after start run - cli uses resolvers fst: "central.sonatype.com ", snd: "repo.scala-lang.org ", low priority: <repo>

Expected behaviour
I expects scala-cli would use <repo> (provided by env COURSIER_REPOSITORIES and command scala-cli --power config repositories.default ) as first priority resolver

[Gedochao]

@geny200 a somewhat odd question - are only Bloop downloads affected? or does this happen with build dependencies and whatnot?

[geny200]

are only Bloop downloads affected?

As far as I understand yes

[Gedochao]

Awesome, then I managed to reproduce it.
As far as I can tell, it is not strictly a regression in v1.15.0 as much as there was a Bloop bump, which may have induced this behaviour where it was cached before.
Either way, fix in progress.