Currently the deployer only regenerates the systemd unit file when the store path changes. If you update env vars or other config in the manifest, the deployer says 'already at' and skips regeneration. It should hash the full service config and regenerate when anything changes.
Simple fix: add a --force flag to ship.sh/push.sh that tells the deployer to regenerate the unit file regardless of store path match. Could work by adding a forceRedeploy field to the manifest service entry, or a separate pendingRedeploys list. Deployer clears the flag after regenerating. This avoids complex config hashing and gives explicit user control.