[BUG] Memory regression in v2.1.62: Claude process consuming ~6.4GB RAM on 6.5GB system, causing repeated OOM kills

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

[Report generated by Claude Code]

Title: Memory regression in v2.1.62: Claude process consuming ~6.4GB RAM on
6.5GB system, causing repeated OOM kills

---
Environment

• Claude Code version: 2.1.62 (upgraded from 2.1.34 on Feb 27 2026)
• OS: ChromeOS Linux container (Crostini), Debian-based
• Architecture: ARM64 (aarch64)
• RAM: 6.5GB total, no swap (Crostini blocks swapon)

Problem

Since upgrading from v2.1.34 → v2.1.62, the Linux kernel's OOM killer
terminates Claude Code repeatedly during use. This did not occur with
v2.1.34 under the same workload.

dmesg shows 4 kills in a single day, each with ~6.4–6.5GB RSS — essentially
the entire available RAM:

Out of memory: Killed process 1515 (claude) anon-rss:6496332kB,
oom_score_adj:200
Out of memory: Killed process 1868 (claude) anon-rss:6487688kB,
oom_score_adj:200
Out of memory: Killed process 2158 (claude) anon-rss:6488628kB,
oom_score_adj:200
Out of memory: Killed process 2354 (claude) anon-rss:6490332kB,
oom_score_adj:200

The new binary is ~10MB larger (224MB vs 215MB), suggesting meaningful
internal changes that may include a memory growth regression.

Contributing factors

• oom_score_adj: 200 — Claude is always the first OOM target
• No swap available in Crostini containers
• Sessions start at ~548MB RSS but grow to 6.4GB before being killed

Expected behaviour

Memory usage comparable to v2.1.34, which did not trigger OOM kills on the
same hardware.

Workaround

Frequent use of /compact and starting new sessions more often partially
mitigates growth but does not prevent it. However process is being killed under very small contexts currently.

What Should Happen?

Memory usage comparable to v2.1.34, which did not trigger OOM kills on the
same hardware.

Error Messages/Logs

Out of memory: Killed process 1515 (claude) anon-rss:6496332kB,
  oom_score_adj:200
  Out of memory: Killed process 1868 (claude) anon-rss:6487688kB,
  oom_score_adj:200
  Out of memory: Killed process 2158 (claude) anon-rss:6488628kB,
  oom_score_adj:200
  Out of memory: Killed process 2354 (claude) anon-rss:6490332kB,
  oom_score_adj:200

Steps to Reproduce

Running almost any Claude Code commands. For example running the diagnostics for this issue completed, but I was unable to respond to Claude asking for an input afterwards.

Claude Model

Sonnet (default)

Is this a regression?

Yes, this worked in a previous version

Last Working Version

2.1.34

Claude Code Version

2.1.62

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

[Report generated by Claude Code]

Title: Memory regression in v2.1.62: Claude process consuming ~6.4GB RAM on
6.5GB system, causing repeated OOM kills

---
Environment

• Claude Code version: 2.1.62 (upgraded from 2.1.34 on Feb 27 2026)
• OS: ChromeOS Linux container (Crostini), Debian-based
• Architecture: ARM64 (aarch64)
• RAM: 6.5GB total, no swap (Crostini blocks swapon)

Problem

Since upgrading from v2.1.34 → v2.1.62, the Linux kernel's OOM killer
terminates Claude Code repeatedly during use. This did not occur with
v2.1.34 under the same workload.

dmesg shows 4 kills in a single day, each with ~6.4–6.5GB RSS — essentially
the entire available RAM:

Out of memory: Killed process 1515 (claude) anon-rss:6496332kB,
oom_score_adj:200
Out of memory: Killed process 1868 (claude) anon-rss:6487688kB,
oom_score_adj:200
Out of memory: Killed process 2158 (claude) anon-rss:6488628kB,
oom_score_adj:200
Out of memory: Killed process 2354 (claude) anon-rss:6490332kB,
oom_score_adj:200

The new binary is ~10MB larger (224MB vs 215MB), suggesting meaningful
internal changes that may include a memory growth regression.

Contributing factors

• oom_score_adj: 200 — Claude is always the first OOM target
• No swap available in Crostini containers
• Sessions start at ~548MB RSS but grow to 6.4GB before being killed

Expected behaviour

Memory usage comparable to v2.1.34, which did not trigger OOM kills on the
same hardware.

Workaround

Frequent use of /compact and starting new sessions more often partially
mitigates growth but does not prevent it. However process is being killed under very small contexts currently.