v1.3.2

This release tightens provider recovery and live reconfiguration, removes several preparation stalls, and makes the resulting work substantially easier to inspect.

Highlights

• Provider circuit breaking is concurrency-aware, recovers promptly from transient failure bursts, and uses a controlled STAT probe to restore providers isolated during BODY preparation.
• Provider graph reloads now drain active clients instead of disposing them underneath in-flight work.
• RAR preparation reuses retained prefixes, bounds expensive parser fallbacks, and avoids redundant full-volume reads for ordinary split archives.
• Watchdog entries persist expandable health-routing and preparation breakdowns, including provider shares, misses, failures, fallback counts, and phase timings.
• Settings metadata, shared validation, and frontend state handling were consolidated; provider testing remains mandatory before save, while connection and health-pipeline benchmark recommendations are shown independently.

Fixes

• Shared-host provider accounts keep independent bandwidth histories through stable provider IDs.
• Queue imports warm eligible provider connections earlier, and health-only providers retain their configured authenticated capacity between jobs.
• Browsers without crypto.randomUUID() can create providers and indexers and regenerate API keys.
• Watchdog timing distinguishes health checks that did not run from completed checks.

Full changelog: v1.3.1...v1.3.2