DynamoDB Adaptive Capacity Split Threshold and Opt-Out Configuration Gap

DynamoDB Adaptive Capacity automatically splits partitions when sustained traffic concentrates on a hot key, yet the service exposes no configuration parameter, API, or console setting to define the throughput threshold that triggers a split, manually initiate one, or opt out of the behavior entirely. For integration tests that must validate partition-boundary assumptions without production credentials, this creates a gap: DynamoDB Local and containerized mocks run with adaptive capacity disabled, so they reflect only the static partition model, while exercising the real service in a development account risks unexpected splits and billing.

The core uncertainty is whether AWS intends to surface split-threshold controls or an opt-out flag for workloads that require deterministic partition behavior during pre-production validation. Without such controls, teams cannot reliably reproduce or freeze the partition topology that their application logic depends on.

Is there a documented configuration boundary or service limit that indirectly constrains when Adaptive Capacity splits partitions? Can Contributor Insights or detailed CloudWatch monitoring provide enough visibility to infer split events without enabling costly high-resolution metrics? Does AWS maintain a public roadmap or feature request channel for adding explicit split-threshold management to DynamoDB tables?