Retry Budgets, Concurrency Bottlenecks, and the Lean Container Push

Tools

PR #315: Retry Budget Caps Added to Scheduler Resilience Layer

This PR tightens the internal retry logic with exponential backoff caps. For anyone running OpenClaw in production, this means fewer surprise cascades when your scheduler hits transient failures.

PR #313: Dynamic Worker Pool Sizing by Default

This bumps the default worker thread pool sizing to scale with available cores instead of a hardcoded value. Should help on beefy CI runners without manual tuning.

PR #312: Leaner Default Container Image with Stripped Language Runtimes

Cuts the default container image by removing unused language runtimes from the base layer. Smaller images mean faster cold starts and lower registry storage costs.

News

Issue #1545: Concurrent Task Scheduling Bottleneck Under High Parallelism

A deep-dive issue documenting CPU throttling during concurrent task orchestration. Useful if you're benchmarking or fighting mysterious latency spikes in multi-pipeline runs.

Issue #1539: Stale Lockfile Left Behind After Unclean Shutdown

A bug report showing that stale lockfiles aren't cleaned up on crash, causing subsequent runs to fail silently. If your pipelines occasionally hang for no reason, this is why.

Analysis

Issue #1540: Proposal to Introduce Build Artifact Fingerprinting

The team is evaluating hash-based caching for intermediate build artifacts. Could cut cold-start times significantly for repeat workflows. Worth tracking if you care about pipeline efficiency.