Max 20x subscriber rate-limited within usage limits during Claude Code subagent orchestration (no early warning / no budget visibility)

Summary

Claude Max 20x subscriber rate-limited during an extended Claude Code session while believing I was within my plan's usage limits, with no early warning that a cap was approaching.

Environment

• Plan: Claude Max 20x
• Client: Claude Code (model claude-fable-5)
• Session type: long-running multi-phase build using heavy subagent orchestration

Usage pattern that triggered it

An extended build in Claude Code that repeatedly fanned out 3–6 concurrent Opus subagents (implementers + reviewers) per phase, running in the background, over several hours. Background subagents appear to draw concurrently from the same token pool as the main loop, so parallel orchestration multiplied consumption much faster than a single-threaded session — and the consumption was not visible anywhere before hitting the wall.

Expected behavior

As a Max 20x subscriber within usage limits, either:

• (a) not be rate-limited at this usage, or
• (b) get a clear, early signal of remaining budget and what is consuming it — especially given that parallel/background subagents make consumption opaque and non-obvious to pace against.

Actual behavior

Hit a rate limit mid-build with no prior warning; the active session stalled. (Work itself was preserved via git commits.)

Impact

Interrupted an active build/demo. The lack of visible budget/consumption for subagent fan-out made it impossible to pace usage before the limit.

Requests

1. Confirm whether this rate-limit trip was correct for a Max 20x plan at this usage.
2. If expected, surface subagent / parallel-agent token consumption and remaining budget in Claude Code so users can pace before hitting the limit (a running "budget remaining" + per-subagent consumption view).