Implements

• Fix for Fallback not triggered when all profiles of a provider timeout #1371

Root cause

• Provider aborts surfaced as AbortError with “request was aborted”, which the fallback classifier treated as user aborts instead of timeouts, so model failover never triggered and runs stopped on the primary.

Fix

• Treat specific provider AbortError messages as timeouts in failover classification so the runner triggers model fallback
• Add e2e to lock the regression.

Checklist

• Added test coverage and tested
• pnpm lint
• Tested and working with my clawdbot instance

Closes #1371