php-cli queue:work segfaults (SIGSEGV) on SIGTERM/SIGINT graceful shutdown (v1.12.0-1.12.4)

What happens

A long-lived Laravel queue:work launched via frankenphp php-cli crashes with SIGSEGV (exit 139, core dumped) when it receives SIGTERM or SIGINT to shut down gracefully. The crash happens at signal-driven shutdown.

Voluntary exit is clean: Laravel's --max-time makes the worker exit 0 every cycle. Only signal-driven shutdown segfaults. This leaves any systemd-managed worker in failed (Result=core-dump) after systemctl stop.

Reproduction

Inside a Laravel app; the queue can be empty (no job in flight):

frankenphp php-cli artisan queue:work --sleep=3 --tries=3 &
P=$!; sleep 4; kill -TERM $P; wait $P; echo $?   # -> 139

Same result with kill -INT.

Versions tested (all segfault)

• v1.12.0 (PHP 8.5.3)
• v1.12.1 (PHP 8.5.3)
• v1.12.4 (PHP 8.5.7, Caddy 2.11.4), static frankenphp-linux-x86_64 — still rc 139 on both TERM and INT

Environment

• Linux x86_64, Ubuntu 24.04, static (musl) binary, systemd unit (non-root user)
• Possibly related to #2342 (ExecuteScriptCLI segfaults)

Can provide a core/backtrace if helpful.

Start with the frankenphp php-cli signal-driven shutdown path and reproduce the issue using the provided Laravel queue:work command with SIGTERM and SIGINT. Compare those results with the clean --max-time exit and review the related ExecuteScriptCLI segfault in issue #2342; done means both signals shut down without SIGSEGV or exit 139.