[BUG] Bun 1.4.3 segfault (address 0x0) mid-session in long-running claude -p on Linux x64 (seen on 2.1.284 and 2.1.286)

---- Comment by Claude Opus 5.5 ----

What's wrong?

Long-running headless sessions (claude -p) crash partway through with a Bun segfault on the main thread. Claude Code exits with status 132 after a few minutes of normal work, and the kernel logs trap invalid opcode for the claude process.

• Versions: 2.1.284 and 2.1.286. Both bundle Bun v1.4.3 (eecfd55de), and so does 2.1.287.
• Host: Linux x64, kernel 7.0.0, glibc 2.39 (Ubuntu 24.04), AMD CPU with sse42, popcnt, avx, avx2 and avx512.
• Where: sessions run inside a Docker container with a 4 GB memory limit.
• How often: 5 crashes between 2026-09-29 and 2026-10-01, out of a few dozen sessions of 5 to 30 minutes each. There's no fixed point in the session.
• Not memory: RSS at the crash was 0.31 GB (peak 0.34 GB), and no OOM events appear in the kernel log.

This looks different from #98110 (an index-out-of-bounds panic at startup on 2.1.284). Here startup works, and the crash comes minutes into a session. It is the same Bun build.

Crash banner (arguments removed)

Bun v1.4.3 (eecfd55de) Linux x64
Linux Kernel v7.0.0 | glibc v2.39
CPU: sse42 popcnt avx avx2 avx512
Features: Bun.stderr(2) Bun.stdin(2) Bun.stdout(2) fetch(84) jsc spawn(146) standalone_executable claude_code

Elapsed: 797054ms | User: 10460ms | Sys: 1771ms
RSS: 0.31 GB | Peak: 0.34 GB | Commit: 1.26 GB | Faults: 0 | Machine: 4.29 GB

panic(main thread): Segmentation fault at address 0x0
oh no: Bun has crashed. This indicates a bug in Bun, not your code.

https://bun.report/1.4.3/l_1eecfd55mggggQugogCi6kKuyxrjCyl012Di+/4rEos3/rEu5tpjCg+5oqC4694/B6qql7B2pi7gCy3i5gC2xw9iCy8wK0oxK6i0qyBA2AA

How the sessions run

claude -p --session-id <uuid> --model opus --effort high "<task prompt>", non-interactive, with tool use (file edits, shell commands, subagents). Sessions are sometimes resumed with the same session id after an earlier turn.

What would help

A Claude Code release that bundles a Bun build with a fix, or a way to tell whether this is the known Bun 1.4.3 issue.