“Sandboxed” can become a dangerously broad trust label. Flatpak isolates applications, but recent vulnerabilities show that isolation has several edges: filesystem path resolution during deployment, privileged helper behavior, and process-group signaling. CVE-2026-97029 is especially useful as a reminder. PID namespace separation did not stop a sandboxed app from signaling unsandboxed processes that shared its process group, potentially terminating components such as the desktop shell. That is not the same impact as reading arbitrary host files or gaining code execution. It is still a boundary failure administrators should understand rather than flatten into a single “sandbox escape” label. **In practical terms, it is a good time to:** - verify the Flatpak build actually installed on managed endpoints - identify distributions using backported fixes rather than relying on upstream version numbers alone - test whether untrusted Flatpak workloads can affect parent-session processes - review endpoint recovery behavior when the desktop shell or session components terminate unexpectedly How often do your security reviews test what a sandbox can influence outside itself, rather than only what it can read or write? #Linux #LinuxSecurity #DesktopSecurity #OpenSource #SecurityOperations https://linuxsecurity.com/features/flatpak-vulnerability-linux-host-files-processes