Supervisor reverts to the first locally pushed version when restarted in local mode

[pipex]

Replication

• Download https://github.com/balena-io-examples/balena-nodejs-hello-world
• Start a local livepush session balena push , wait for the device to settle
• Stop the livepush session, and make a small code change
• Start a new live push, check that your changes are applied. You'll also see the image id change
• Restart the supervisor, the supervisor will log something like this

Aug 27 18:31:50 68568ed balena-supervisor[10222]: [debug]   Replacing container for service balena-hello-world because of config changes:
Aug 27 18:31:50 68568ed balena-supervisor[10222]: [debug]     Non-array fields:  {"added":{},"deleted":{},"updated":{"image":"sha256:dda4ef12f2dd49f1d7a8c7fa2f2ca35a132a43c65380893737a531a80ea9027b"}}
Aug 27 18:31:50 68568ed balena-supervisor[10222]: [debug]   Replacing container for service balena-hello-world because of config changes:
Aug 27 18:31:50 68568ed balena-supervisor[10222]: [debug]     Non-array fields:  {"added":{},"deleted":{},"updated":{"image":"sha256:dda4ef12f2dd49f1d7a8c7fa2f2ca35a132a43c65380893737a531a80ea9027b"}}

The problem is that the supervisor only generates a saveImage step if the image tag does not exist in the database, we need to compare that the image exists on the database with the same docker id and tag here

balena-supervisor/src/compose/application-manager.ts

Line 640 in 3049b03

function saveAndRemoveImages(

This does not affect cloud releases because a change always results in a new image tag, while in local releases, the image tag is re-used for every push