Bug: tmux coordinator fails to retrieve PID in containerized environments (Fedora Silverblue/Toolbox)

[GalegO]

Steps to reproduce

Description:

When running Overstory inside a Fedora Silverblue Toolbox (or similar OCI containers), the ov coordinator start command fails to retrieve the PID of the tmux session. This appears to be caused by how Overstory manages the tmux socket path, which conflicts with the container's isolated /tmp namespace and permission model.
Steps to Reproduce:

Use Fedora Silverblue with a standard Toolbox/Distrobox container.

Install overstory-cli and tmux inside the container.

Run ov init and git init.

Attempt to start the coordinator: ov coordinator start.

Expected behavior

Expected Behavior:

The coordinator should initialize the tmux session, retrieve the PID, and remain running in the background.

Actual behavior

Actual Behavior:

The command fails with the following error:
Error [AGENT_ERROR]: Created tmux session "overstory-test-coordinator" but failed to retrieve PID: no server running on /tmp/tmux-1000/overstory

Even if the session is manually verified via tmux ls, the Overstory CLI cannot communicate with it, likely due to the socket being inaccessible or incorrectly mapped between the container and the host's /tmp.

tmux version:

tmux 3.5a

overstory version

0.9.3

Bun version

1.3.12

Operating system

Fedora Silverblue 43