Create a new corpus. Perform some API search (e.g. OpenAlex) to fill in docs. If the API fails for some reason, the job is terminated and one gets partially imported docs but no terms in the table.

This should be fixed, so that even despite the API failing (and a proper error message being shown to the user), the job can continue with filling in terms etc.

Also, it would be nice to find out exactly which code is responsible for this. I think it's somewhere in the Flow.hs file, but I'm not exactly sure where.