@jeroen.dries suggested to enable the “soft error” feature, so that a single failing file doesn’t ruin the whole job.
In Python it can be enabled for a batch job with
data_cube.execute_batch(job_options={'soft-errors': 'true'})
But I don’t known how that can be done with the R client.
job_options also not standardized yet, so it might not be possible at all (yet) with the R client.