GitHub issue #68619 in anthropics/claude-code, opened 2026-06-15 by lonlundgren, records a report that subagents recursively spawn child agents 50+ levels deep while ignoring the CLAUDE_CODE_FORK_SUBAGENT=0 setting, burning 1.2M+ tokens in ~30 minutes on a task that should have been a git clone; the report characterizes the behavior as a major regression shipped on or around June 10th.
Verbatim Excerpts from Cited Records
“Subagents recursively spawn child agents 50+ levels deep, ignoring `CLAUDE_CODE_FORK_SUBAGENT=0`.”
·
anthropics/claude-code (issue 68619 opening body, Summary section)
·
2026-06-15T15:41:23Z
·GitHub Issue ↗
“In the observed case: 1.2M+ tokens consumed in ~30 minutes on a task that should have been `git clone` + `find . -name '*.sol'`.”
·
anthropics/claude-code (issue 68619 opening body, Summary section)
·
2026-06-15T15:41:23Z
·GitHub Issue ↗
“This is a **major regression that shipped on or around June 10th**.”
·
anthropics/claude-code (issue 68619 opening body, Summary section)
·
2026-06-15T15:41:23Z
·GitHub Issue ↗
Counter-record: No in-thread refutation of the reported recursive-spawning behavior as of 2026-10-10: 35 comments corroborate the runaway subagent tree and token burn. The report documents six compound bugs (FORK_SUBAGENT=0 ignored, permission denial triggering recursion, no permission propagation, HTTP per-file fetching instead of cloning, lost intermediate work on interruption, retry storms) and a chain analysis of how they compound.
Falsification probe: If issue #68619 is deleted without archive, if the author edits the opening body so the quoted sentences no longer match verbatim, or if the author/maintainers post a definitive retraction showing the behavior was misattributed, this claim drops to Withdrawn.
Audit verification: Verified 100% continuous verbatim substring against official GitHub REST API by
scripts/verify_claim_verbatim.py.