Skip to content

Instantly share code, notes, and snippets.

@LEX8888

LEX8888/claude-instability-report-2026.md

Last active February 3, 2026 21:46
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save LEX8888/675867b7f130b7ad614905c9dd86b57a to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save LEX8888/675867b7f130b7ad614905c9dd86b57a to your computer and use it in GitHub Desktop.
Download ZIP
Claude Desktop/Code: 19 incidents in 14 days, memory leak shipped to production (Jan 27 - Feb 3, 2026)
Raw
claude-instability-report-2026.md

Claude Desktop/Code: A Week of Instability (Jan 27 - Feb 3, 2026)

Executive Summary

In the span of just one week, Anthropic's Claude services experienced unprecedented instability: 19 official incidents logged on their status page, 1,469 GitHub issues opened, and a critical memory leak bug shipped to production that caused systems to crash within 20 seconds of use.

📊 Key Statistics

Metric Value
GitHub Issues (1 week) 1,469
Bug-labeled Issues (1 week) 885
500 Error Issues 105
Performance Issues 43
Official Incidents (2 weeks) 19
Average Incidents per Day 1.36

Claimed vs Calculated Uptime

Service Claimed Uptime (90 days)
claude.ai 99.41%
platform.claude.com 99.82%
Claude API 99.68%
Claude Code 99.70%

Actual Downtime Calculation (Jan 20 - Feb 3, 2026)

Date Incident Duration (min)
Feb 3 API all models 15
Feb 2 Opus 4.5 6
Feb 2 Credit purchase issues ~360 (6h)
Feb 1 Sonnet 4.5 19
Feb 1 Opus 4.5 10
Jan 31 Memory leak 2.1.27 ~1,440 (24h)
Jan 29 Credits delay ~90
Jan 28 Opus 4.5 (x2) 71
Jan 28 Sonnet 3.7 23
Jan 27 Console degraded 39
Jan 27 Haiku 3.5 33
Jan 26 Opus 4.5 ~360 (6h)
Jan 23 Signature errors ~480 (8h)
Jan 22 Multiple incidents ~60
Jan 21 Sonnet 4.5 48
Jan 20 Opus + Sonnet 53
TOTAL ~3,207 min (~53 hours)

The Math

  • 14 days = 336 hours = 20,160 minutes
  • Total downtime: 3,207 minutes
  • Calculated uptime: (20,160 - 3,207) / 20,160 = 84.1%
Metric Value
Claimed uptime 99.41%
Calculated uptime ~84.1%
Discrepancy ~15%

Note: Even excluding the 24-hour memory leak (software bug vs infrastructure), the calculated uptime is ~91.4% — still far below claimed 99.41%.

19 incidents in 14 days = 9.5 incidents per week

For comparison:

  • OpenAI typically reports 1-2 incidents per week
  • Google Cloud typically reports 2-3 incidents per week

🔴 Official Incidents Timeline (status.claude.com)

February 3, 2026

Elevated error rate on API across all Claude models

February 2, 2026

Elevated errors on Claude Opus 4.5

Credit purchase issues & delays

February 1, 2026

Elevated errors on Claude Sonnet 4.5

Elevated errors on Claude Opus 4.5

January 31, 2026

Memory leaks in Claude Code 2.1.27

January 29, 2026

API credits purchase delays

January 28, 2026

Elevated errors on Claude Opus 4.5 (TWO separate incidents)

Elevated errors on Claude Sonnet 3.7

January 27, 2026

Degraded performance on Claude Console

Elevated errors on Claude Haiku 3.5

January 26, 2026

Increased rate of errors for Opus 4.5

January 23, 2026

Elevated errors: Invalid signature in thinking block

January 22, 2026

Elevated error rates on claude.ai and API + Claude Code Authentication Issues + Elevated errors on claude.ai

January 21, 2026

Elevated errors on Claude Sonnet 4.5

January 20, 2026

Elevated errors on Claude Opus 4.5 and Sonnet 4.5

🐛 Top GitHub Issues by User Reactions

#1: Critical Memory Regression (👍90 reactions, 59 comments)

[BUG] Critical memory regression in 2.1.27 - OOM crash on simple input

  • anthropics/claude-code#22042
  • Date: January 30, 2026
  • Problem: Memory consumption explodes from 467MB to 7.5GB within 20 seconds
  • Status: CLOSED (fixed in 2.1.29)
| Version | Behavior after typing "Test" |
|---------|------------------------------|
| 2.1.27  | 467MB → 7.5GB in 20 sec → OOM killed |
| 2.1.25  | ~540-555MB, stable |

OOM Killer Log Evidence:

[Fri Jan 30 15:40:20 2026] Out of memory: Killed process 2629268 (claude) 
total-vm:86031684kB, anon-rss:7510204kB

#2: Freeze Regression (👍45 reactions)

[BUG] Claude Code 2.1.27 freezes after first message - regression

#3: CPU Hang (👍27 reactions)

v2.1.27 hangs at 100% CPU after first response (v2.1.25 works)

#4: MCP Tool Calls Broken (👍26 reactions)

Claude Desktop: MCP tool calls shown in UI but never sent to server

#5: Quality Regression (👍11 reactions)

Massive quality regression

User Quote:

"Since 26.01.2026 Claude code started working just disgustingly. It makes multiple broken attempts instead of thinking through the problem. It started thinking much less. So I even don't want to use it anymore. It's gone stupid. Please return things to the way they were."

💬 User Testimonials (Hacker News, February 3, 2026)

On Infrastructure Stability:

"Anthropic might have the best product for coding but good god the experience is awful. Random limits where you know you shouldn't hit them yet, the jankiness of their client, the service being down semi-frequently. Feels like the whole infra is built on a house of cards and badly struggles 70% of the time. I think my $20 OpenAI sub gets me more tokens than Claude's $100."

On Comparison with Competitors:

"I've had the $20/month account for OpenAI, Google, and Anthropic for months. Anthropic consistently has more downtime and throws more errors than the other two. Claude (on the web) also has a lot of seemingly false positive errors. It will claim an error occurred but then work normally."

On Subscription Cancellations:

"I've cancelled my CC subscription, I'm not paying to support anticompetitive behaviour."

"I kind of lost interest in local models. Then Anthropic started saying I'm not allowed to use my Claude Code subscription with my preferred tools and it reminded me why we need to support open tools and models."

On Account Bans:

"Anthropic banned my account when I whipped up a solution to control Claude Code running on my Mac from my phone when I'm out and about. No commercial angle, just a tool I made for myself since they wouldn't ship this feature."

On Pro vs Max Tiers:

"Even when Claude isn't down on status indicator. I get issues with Claude all the time being very slow where it is basically down. I have to kill the request and then restart it. I wonder if it has to do with being on Pro plan. I rarely had this issue when I was on Max."

Anthropic's Official Response (Feb 3, 2026):

"Hey folks, I'm Alex from the reliability team at Anthropic. We're sorry for the downtime and we've posted a mini retrospective on our status page. We're also be doing a more in depth retrospective in the following days."

📈 Problem Timeline

Date Event
Jan 20 Elevated errors on Opus 4.5 and Sonnet 4.5
Jan 21 Elevated errors on Sonnet 4.5
Jan 22 Multiple incidents: auth issues, elevated errors
Jan 23 Invalid signature errors (~8 hours)
Jan 26 Quality regression complaints begin
Jan 27 Harness bug in Claude Code (context 65% vs 98%)
Jan 27 Console degradation, Haiku errors
Jan 28 Issue "Massive quality regression" opened
Jan 28 THREE separate incidents (Opus x2, Sonnet)
Jan 30 Version 2.1.27 released with critical memory leak
Jan 30 Multiple incidents
Jan 31 Flood of freeze/hang issues for 2.1.27
Jan 31 Status page acknowledges memory leak
Feb 1 Elevated errors Opus 4.5, Sonnet 4.5
Feb 2 Elevated errors Opus 4.5, credit purchase issues
Feb 3 Major outage 15 minutes, 128 issues in 3 hours

🔗 Key References

Status Page

GitHub Issues

Hacker News

External Monitoring

📝 Key Findings

  1. Version 2.1.27 was released with a critical memory leak that caused OOM crashes within 20 seconds of use. This passed QA and reached production users.

  2. 19 official incidents in 14 days - approximately 1.36 incidents per day, or 9.5 per week. This is 3-5x higher than typical rates for OpenAI or Google Cloud.

  3. Users report degraded quality since January 26 - the model makes "multiple broken attempts instead of thinking through the problem."

  4. No compensation offered - despite multiple outages and a shipped bug that made the product unusable, no credits or refunds have been mentioned in any incident report.

  5. Pro tier users report worse experience than Max tier users, suggesting possible resource prioritization or throttling.

  6. 5,788 open issues on the Claude Code GitHub repository as of February 3, 2026.

Conclusion

While Anthropic claims 99.41% uptime on their status page, the user experience tells a different story. With 19 incidents in two weeks, a production-shipped memory leak, and widespread user complaints about quality degradation and infrastructure instability, paying customers are receiving significantly degraded service with no compensation.

As one Hacker News user summarized: "Feels like the whole infra is built on a house of cards and badly struggles 70% of the time."

Data collected: February 3, 2026, 18:00 UTC Sources: status.claude.com, GitHub, Hacker News

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment