I loaded a very small site-only VCF (200k sites), annotated it with VEP and then export an annotated VCF (unfortunately my collaborators only work with VCF). The process took 7.5 hours. Is that something we expect?
The slow thing here isn’t exporting the VCF, it’s VEP. Each VEP process can annotate on the order of 10 variants per second – what’s your cluster configuration here?