This is not a bug; the feature is there to allow users to work on the repositories, without affecting repo consumers. And once the repo is ready, regenerate the repo for all the new builds (and thus ship all the RPMs at once).

In that case the bug is in the fields description.

Repository meta data is normally refreshed after each build. If you want to do this manually, turn this option on.

Change to (something like):

Repository meta data is normally refreshed and published after each build. If you want to publish metadata manually, turn this option on. Note that builds in this copr will always see refreshed metadata regardless of this option.

Ack, last /usr/bin/createrepo_c --database --ignore-lock --local-sqlite --cachedir /tmp/ --workers 8 --update /var/lib/copr/public_html/results/@python/python3.8/fedora-rawhide-x86_64 run took more than 7 minutes.

+1 to this RFE. In my use case (iucar/cran), I generate the dependency tree, and then create a list of build batches, so that there are no interdependencies in a batch. This is actually needed, because Copr does not guarantee that any build will wait for any other (and that's fine). But then we should be able to

1. submit a batch of packages to build, so that none of them calls createrepo;
2. call createrepo when the batch finishes;
3. repeat.

In this way, resources are not wasted. There could be a separate option to publish or not the result from createrepo. But disabling createrepo altogether is required too (and that's what the current wording suggests).

Note that this also requires some way to call createrepo programatically, instead of the button in the web UI (see #960).

+1 to this RFE

This is only documentation bug, the build-batch feature is not yet filled in this tracker;
feel free to submit one.

This issue has been migrated to GitHub: https://github.com/fedora-copr/copr/issues/908