[BUG] Intermittent "Subprocess initialization did not complete within 60000ms" on macOS — interactive sessions (VS Code extension / desktop app / interactive CLI) hang, `claude -p` works; correlates with TLS-inspection (Norton)
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?
Intermittent error on startup: Subprocess initialization did not complete within 60000ms — check authentication and network connectivity.
Affects interactive sessions — the VS Code extension, the Claude desktop app, and interactive claude started in a git repo — which hang for exactly 60s. claude -p "say ok" generally succeeds, and subcommands (--version, doctor, plugin list, mcp list) always work.
Ruled out (each tested in isolation; still hung):
- Plugins — uninstalled the only one; logs show
Found 0 plugins - MCP —
ENABLE_CLAUDEAI_MCP_SERVERS=false; removed project.mcp.json - Telemetry / feature flags —
DISABLE_TELEMETRY=1,CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 - Remote Control —
disableRemoteControl: true(user + project) - Git performance —
git status~0.03s on the machine - Network — reproduced on a phone hotspot, off the corporate network
- Auth —
claude doctorshows OAuth present / subscriber auth active - Full
rm -rf ~/.claude+ reinstall, andclaude update— each only helped briefly before recurrence
Diagnostic evidence:
- DEBUG log consistently ends at
Found 0 plugins/ git-remote parse, then emits nothing for 60s until the timeout. sampleof the hung process: main thread idle inkevent64, all workers in__psynch_cvwait— waiting on an async network response that never arrives (not CPU, not a file lock, not a child process).lsofwhile hung: theclaudeprocess holds ESTABLISHED TCP connections to Anthropic (160.79.104.10:443) and a Google-cloud host (34.149.66.x:443) but receives no response.claude doctor:Remote Control: enabled;tengu_ccr_bridge gate (no server response this session).- Trigger: an empty directory +
git remote add origin <any URL>reproduces the hang; an empty dir with no remote starts instantly.
Likely cause: the machine runs Norton with an active network extension (com.norton.mes.networkextension) performing TLS interception — a Norton trusted CA root certificate is installed in /Library/Keychains/System.keychain and Claude Code loads it ("1 system CA certificate" in debug). The inspection proxy establishes the startup connection but does not pass the HTTP/2 / streaming request-response through, so the startup probe waits out the full 60s. Disabling Norton Intrusion Prevention makes the CLI start instantly; with it on (even with domain exclusions) the timeout returns intermittently.
Related issues: #50988, #65440, #18666, #57556, #50559.
What Should Happen?
Startup should fail fast with an actionable message when the startup network probe can't complete (e.g. behind a TLS-inspecting proxy/AV), instead of hanging for the full 60s. Ideally provide an offline / skip-probe startup path, and document TLS-inspection / corporate-proxy incompatibility.
Note: the main API connection to api.anthropic.com works fine — only this startup probe stalls under TLS inspection. A shorter timeout + clear error (or a documented way to bypass the probe) would unblock affected machines.
Error Messages/Logs
Subprocess initialization did not complete within 60000ms — check authentication and network connectivity
DEBUG log (trimmed) — init proceeds normally, then stops dead with no further output until the 60s timeout:
[init] configureGlobalMTLS complete
CA certs: stores=bundled,system
CA certs: Loaded 145 bundled root certificates
CA certs: Loaded 1 system CA certificates <- the Norton MITM CA
mTLS: Creating HTTPS agent with custom certificates
...
Found 0 plugins (0 enabled, 0 disabled)
<- ~60s of nothing, then the timeout error above
claude doctor:
Remote Control: enabled
tengu_ccr_bridge gate (no server response this session)
(Raw logs withheld here as they contain private file paths; can be shared privately on request.)
Steps to Reproduce
- On macOS with an active TLS-inspection security agent (here: Norton — system extension
com.norton.mes.networkextension, which installs aNorton trusted CAroot cert in the System keychain). - In an EMPTY directory, run
claude -p "say ok"— starts/exits instantly (works). - Make it a git repo with a remote:
````
mkdir /tmp/repro && cd /tmp/repro && git init && git remote add origin https://github.com/any/repo.git
- Open the Claude Code VS Code extension in that folder, OR run interactive
claudethere. - Result: startup hangs ~60s, then errors with "Subprocess initialization did not complete within 60000ms".
claude -p "say ok"in the same folder usually still succeeds. - Turn off Norton Intrusion Prevention (the TLS-inspection layer) → startup is instant, confirming the interception is the trigger.
Minimal trigger: any git repo with a remote, on a machine performing TLS interception. An empty dir with no remote does not reproduce.
Claude Model
None
Is this a regression?
I don't know
Last Working Version
_No response_
Claude Code Version
2.1.169 (Claude Code) — also reproduced on 2.1.153 and 2.1.178
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
- Also affects the Claude desktop app's sessions; the Cowork mode in the desktop app keeps working (likely a different network path), which is the only reliable surface during the outage.
- Reinstalling/updating Claude Code and a full
rm -rf ~/.claudeonly resolved it briefly before recurrence — consistent with an environmental (TLS-inspection) trigger rather than a config/version one. - Likely affects any developer machine running Norton (or similar TLS-inspecting endpoint security), so this may be a fleet-level concern for orgs that deploy such tools.
- Happy to provide raw debug logs and the
sample/lsofoutput privately (withheld here as they contain private file paths).
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗