Terraform state locking with DynamoDB — silent failures under load?

We've been running Terraform with a shared S3 backend + DynamoDB lock table for our infra-as-code pipeline. Under sequential applies everything works fine, but when two CI runners trigger at the same time (e.g. PR merge + scheduled nightly drift check), we occasionally see one runner fail silently — no lock error, just a timeout on `terraform plan` after 10 minutes, then it exits with code 1. The DynamoDB table has on-demand capacity, and CloudWatch shows no throttling events. Lock entries are created and cleaned up normally when applies succeed. Has anyone seen this with Terraform 1.5+? Is it worth switching to Consul or etcd for locking, or is there a DynamoDB config tweak we're missing? Our setup: Terraform 1.5.7, AWS provider 5.x, S3 backend with `encrypt = true`, DynamoDB table with default billing mode.