Container fails to restart with `--local` due to non-idempotent Soroban config upgrade

[leighmcculloch]

What version are you using?

quickstart: fbf1664 (latest main)

What did you do?

1. Started the quickstart container with --local
2. Let the container run for a while (services came up successfully)
3. Stopped the container (without removing it)
4. Started the same container again

What did you expect to see?

The container should restart successfully and all services should come back up.

What did you see instead?

The container fails during the Soroban config upgrade process with the following errors:

upgrades: soroban config: restore contract: bdd3c5993d521567bdafeb91a2e93603feba72bfcc9b3e40f9aca27c2aae23d7 .. ERROR
upgrades: soroban config: install contract: 5f9144833e1b8586f05bfb32984c06065b5df1930b7f4f234cb0f910497ecd39 .. ERROR
upgrades: soroban config: deploy contract: e428981f98b699522518f9f744aed1f7117a9fdb444bfb72e6b68e142bc479d4 .. ERROR
upgrades: soroban config: upload config: 8f1eac96d197adbfacf5f2df42dbb81bf24f78ce6ff1950f3f20fd7a09f3c2ff .. ERROR
Error setting configUpgradeSet
!!!!! Unable to upgrade Soroban Config Settings. Stopping all services. !!!!!

All services are then shut down via SIGQUIT.

[leighmcculloch]

The upgrade_local() function in the start script unconditionally calls upgrade_soroban_config() for protocol versions >= 20. When the container is restarted (stopped then started without removing), the ledger data persists within the container's filesystem. The Soroban config upgrades were already applied during the first run, so attempting to re-apply them causes all upgrade transactions to fail.

The script detects the failure via the "Error setting configUpgradeSet" response and calls kill_supervisor to shut down all services.