Feature request: auto-retry on API 500 errors

Problem

When the Anthropic API returns 500 errors during an active incident, Claude Code hangs silently with no visual feedback and no automatic retry. The user sees no spinner, no error message, no indication that anything went wrong — it just looks like Claude is frozen.

The only recourse is to manually press Escape and re-trigger the request, which the user has to figure out on their own.

Requested Behavior

1. Auto-retry with backoff — On 500/502/503 errors, retry 2-3 times with exponential backoff before surfacing the error to the user.
2. Visible status indicator — Show something in the CLI when a request fails and is being retried (e.g., "API error, retrying (1/3)...").
3. Configurable — Allow users to set retry count and backoff in settings (e.g., api.retryCount: 3, api.retryBackoffMs: 2000).

Context

During the Feb 25, 2026 "Elevated error rates across multiple models" incident (status.claude.com), Claude Code became effectively unusable — repeated silent hangs with no feedback. The user experience was frustrating because there was zero indication of what was happening.

This is especially impactful for users running long multi-step tasks where a single 500 error mid-flow kills momentum.