A big portion of the jobs data payload is represented by the repetition of reference data. We should at least consider to retrieve the reference data in one (or several) requests and join it in the ui once the job request has completed. I would expect improvements in data transfer time and memory consumption.

After discussion we don't think this is viable, since many of the reference data tables have too many rows to be able to return all of them to the client (eg job_type has 33K rows, reference_data_signatures 340K rows, machine millions of rows) - and we can't easily work out which will be relevant to the jobs requested next.