Hi everyone I am seeing inconsistent behavior with DeepVariant. • Works fine on GIAB data • On some datasets, DeepVariant shards take extremely long (up to 24h) or appear to stall • Running on GCP (Google Batch), with intervals (~20+ shards), using 16 CPUs What’s odd: the actual variant calling step seems fast, but total runtime is very long → possibly infrastructure-related? Has anyone experienced this or has tips to optimize DeepVariant performance in Sarek? Thanks 🙏