At the moment, when scheduling a job through the API, new CIVs are created for all inputs, even if an existing image is provided as input. The behavior when scheduling a job should match what we do on the "Try out algorithm" page, where we reuse CIVs when possible.
I might give this a try this week, although it is a short week with DIAG day on Wednesday and the AWS workshop on Thursday, so might not get to making a PR.
At the moment, when scheduling a job through the API, new CIVs are created for all inputs, even if an existing image is provided as input. The behavior when scheduling a job should match what we do on the "Try out algorithm" page, where we reuse CIVs when possible.
This relates to https://github.com/comic/grand-challenge.org/issues/3139 - both require a heavy refactoring of civ creation and validation around algorithm jobs.