Preserve leftover recordings on container restart

[chalapathi444]

What

The entrypoint wipes /home/egress/tmp on every start. If a container restarts mid-egress (crash, OOM, redeploy), the in-progress recording in tmp is deleted before anyone can recover it.

How

Before the cleanup, if tmp is non-empty, copy its contents to /home/egress/recovered so operators can retrieve partial output from the previous session:

if [ -d /home/egress/tmp ] && [ "$(ls -A /home/egress/tmp 2>/dev/null)" ]; then
    mkdir -p /home/egress/recovered
    cp -r /home/egress/tmp/. /home/egress/recovered/
fi
rm -rf /home/egress/tmp/*

Notes

• When tmp is empty (the normal case), nothing is copied and startup is unchanged.
• recovered/ is not auto-pruned — open to a retention/size cap if you'd prefer one (e.g. clearing it after N restarts or above a size threshold), happy to adjust.

[CLAassistant]

All committers have signed the CLA.

[biglittlebigben]

Thanks for the suggestion. We do not think that copying files within the pod is a good idea: this could cause the pod disk usage to grow without limit and no process to cleanup. This also doesn't cover Amy of the cases where the pod is killed and restarted from a clean container.