Summary
Some consistency test suites are heavy enough that even after better test selection and matrix reduction, they may still dominate CI wall-clock time.
CI-level sharding could help reduce end-to-end latency for these suites.
Motivation
Recent CI analysis suggests that a small number of heavy directories / files dominate runtime. Structural improvements should come first, but sharding remains an important lever for overall throughput.
Proposed direction
Introduce stable sharding for heavy consistency suites, for example by:
- directory-level splits
- explicit file groups
- stable test-group shards chosen for balanced wall-clock time
This should be done in a way that keeps failures understandable and avoids excessive maintenance burden.
Scope
Focus on the heaviest consistency suites first, then expand if useful.
Acceptance criteria
- at least one heavy consistency job is split into multiple balanced shards
- overall PR CI wall-clock time decreases
- shard boundaries remain understandable and maintainable
Authored by OpenClaw (model: gpt-5.4)
Summary
Some consistency test suites are heavy enough that even after better test selection and matrix reduction, they may still dominate CI wall-clock time.
CI-level sharding could help reduce end-to-end latency for these suites.
Motivation
Recent CI analysis suggests that a small number of heavy directories / files dominate runtime. Structural improvements should come first, but sharding remains an important lever for overall throughput.
Proposed direction
Introduce stable sharding for heavy consistency suites, for example by:
This should be done in a way that keeps failures understandable and avoids excessive maintenance burden.
Scope
Focus on the heaviest consistency suites first, then expand if useful.
Acceptance criteria
Authored by OpenClaw (model: gpt-5.4)