Issue Station
Sign In
Tools hereToolbox

claude-code - 💡(How to fix) Fix [Service Disruption] Max 20x window burned with no productive output — rate-limit failures Apr 28–29 2026 [2 comments, 3 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
anthropics/claude-code#54726•Fetched 2026-04-30 06:37:43
View on GitHub
Comments
2
Participants
3
Timeline
6
Reactions
0
Author
gofullthrottlegofullthrottle
Participants
github-actions[bot]github-actions[bot]gofullthrottlegofullthrottleRickDavis404RickDavis404
Timeline (top)
labeled ×3commented ×2cross-referenced ×1

Filing as a paid Max 20x subscriber experiencing the same window-burn pattern that was acknowledged in the Apr 23 engineering postmortem, recurring through Apr 28–29 alongside the Apr 28 platform outage.

Root Cause

Filing as a paid Max 20x subscriber experiencing the same window-burn pattern that was acknowledged in the Apr 23 engineering postmortem, recurring through Apr 28–29 alongside the Apr 28 platform outage.

RAW_BUFFERClick to expand / collapse

Summary

Filing as a paid Max 20x subscriber experiencing the same window-burn pattern that was acknowledged in the Apr 23 engineering postmortem, recurring through Apr 28–29 alongside the Apr 28 platform outage.

Symptoms

  • Current 5-hour session window: ~90% consumed, effectively zero productive output.
  • Repeated 429 rate-limit responses on Claude Code retries.
  • Failed Deep Research jobs that consumed quota without returning completed reports.
  • Pattern matches the March 26 thinking-history token-burn bug described in the Apr 23 postmortem (idle-session thinking history was supposed to clear once but cleared every turn, dramatically inflating token usage).

Environment

  • Plan: Max 20x
  • Claude Code: v2.1.117 (post-fix per the Apr 23 postmortem)
  • OS: macOS (Darwin 24.6.0)
  • Date of impact: 2026-04-28 → 2026-04-29
  • Overlapping with: the Apr 28 ~17:41–19:00 UTC major outage on status.anthropic.com

Why I'm filing here

The Apr 23 announcement noted three bugs affecting Claude Code, the Agent SDK, and Cowork, and Anthropic reset usage limits for all subscribers as compensation. The current burn behavior on v2.1.117 looks indistinguishable from the pre-fix pattern, at least for users running Deep Research–heavy workloads through Claude Code sessions. Flagging here so engineering has a current-version data point.

What would help

  • Confirmation of whether residual symptoms post-Apr 23 fix are expected, or whether a new regression has been observed.
  • Guidance on capturing diagnostics that would be useful (request IDs, anthropic-ratelimit-* headers, transcript artifacts) — happy to attach.
  • For account-specific credit, I have a separate ticket open with support.

For other affected users

If you're seeing the same thing, please add a comment with: plan tier, Claude Code version, approximate timestamps (UTC), and whether the workload was Claude Code, Deep Research, or both. Aggregated data will help engineering scope the residual issue.

Mark as duplicate if there's an existing megathread for this window — I searched and didn't find one, but happy to consolidate.

extent analysis

TL;DR

The user may need to capture diagnostics, such as request IDs and anthropic-ratelimit-* headers, to help engineering investigate the residual symptoms of the window-burn pattern.

Guidance

  • Capture diagnostics: Collect request IDs, anthropic-ratelimit-* headers, and transcript artifacts to provide useful data for engineering.
  • Verify symptoms: Confirm whether the observed behavior matches the pre-fix pattern described in the Apr 23 postmortem.
  • Check for similar issues: Look for existing megathreads or issues related to the window-burn pattern and Claude Code version 2.1.117.
  • Provide additional information: Share plan tier, Claude Code version, approximate timestamps (UTC), and workload type (Claude Code, Deep Research, or both) to help engineering scope the residual issue.

Notes

The issue seems to be related to a previously acknowledged bug, and the user is experiencing similar symptoms despite being on a post-fix version of Claude Code. The user has already opened a separate ticket with support for account-specific credit.

Recommendation

Apply workaround: Capture diagnostics and provide additional information to help engineering investigate the residual symptoms, as the issue may be related to a new regression or incomplete fix.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#orchestration issue#cache issue#memory leak#API versioning#request timeout

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

claude-code - 💡(How to fix) Fix [Service Disruption] Max 20x window burned with no productive output — rate-limit failures Apr 28–29 2026 [2 comments, 3 participants]