@ttomecek This is already implemented and has been for quite some time (two months at least). If you resubmit the same SCMURL, it will attempt to build the failed components again, using whatever was built in the previous attempt. I have indeed utilized this feature several times when building base-runtime, dealing with randomly failing builds and flaky test suites.

If you update the input and submit a new SCMURL, this won't happen, however -- it will reuse components from the last successful build instead, not the partially built one.

it will reuse components from the last successful build instead

and this is what I'm talking about here -- I suggest reusing from failed module builds as well

I can submit patch for this since it's 2 characters change.

Example of how this is slowing us is here: libtool is one of the first packages to build in shared-userspace -- it was build from same scmurl, twice:

https://koji.fedoraproject.org/koji/taskinfo?taskID=19490674
https://koji.fedoraproject.org/koji/taskinfo?taskID=19482769

I thing the latter build shouldn't have happened and former build should have been reused.

I am going to close this issue based on an IRC discussion. Resolution is:

if a module contains packages which fail during build in random, the module author should NOT change the modulemd and perform a rebuild -- changing modulemd results into rebuilding the whole module

This issue has been migrated to Fedora Forge:
https://forge.fedoraproject.org/koji/mbs/issues/549

Please continue any further discussion there.