bug: AppConfig StopDeployment returns ROLLED_BACK but deployment readback later becomes COMPLETE

[AndreKurait]

Reproduced on unmodified localstack/localstack-pro:dev, freshly pulled on 2026-10-10, native linux/amd64:

• Manifest sha256:513f218caf22d1fcb115f055225c96c7140813efcb23e48c9e24b688cdb31cd4
• Image sha256:b104c9b43442fba4e1e00ab68c447681b08db6dc6128273d105e1db5c87e5ab3
• Version 2026.10.0.dev97; SERVICES=appconfig; no extensions, source mounts, or custom entrypoint.

StopDeployment returns State=ROLLED_BACK, but GetDeployment and ListDeployments continue to report BAKING and later report COMPLETE. The stop does not remain reflected in the observable deployment state.

Observed readback after the stop:

Elapsed	GetDeployment	ListDeployments
0 seconds	BAKING	BAKING
1 second	BAKING	BAKING
5 seconds	BAKING	BAKING
35 seconds	COMPLETE	COMPLETE

Expected: after a successful rollback, the same deployment's read APIs should converge to the rolled-back state, rather than completing the canceled deployment.

Minimal reproduction against a disposable LocalStack instance:

import time
import uuid
import boto3
from botocore.config import Config

client = boto3.client(
    "appconfig", endpoint_url="http://localhost:4566", region_name="us-east-1",
    aws_access_key_id="000000000000", aws_secret_access_key="test",
    config=Config(retries={"max_attempts": 0}),
)
name = "appconfig-repro-" + uuid.uuid4().hex[:8]
app = client.create_application(Name=name)["Id"]
profile = client.create_configuration_profile(
    ApplicationId=app, Name="config", LocationUri="hosted", Type="AWS.Freeform",
)["Id"]
env = client.create_environment(ApplicationId=app, Name="test")["Id"]
version = client.create_hosted_configuration_version(
    ApplicationId=app, ConfigurationProfileId=profile,
    Content=b'{"v":1}', ContentType="application/json",
)["VersionNumber"]
strategy = client.create_deployment_strategy(
    Name=name, DeploymentDurationInMinutes=1, GrowthFactor=50,
    FinalBakeTimeInMinutes=0, ReplicateTo="NONE",
)["Id"]
started = client.start_deployment(
    ApplicationId=app, EnvironmentId=env, ConfigurationProfileId=profile,
    ConfigurationVersion=str(version), DeploymentStrategyId=strategy,
)
scope = dict(ApplicationId=app, EnvironmentId=env,
             DeploymentNumber=started["DeploymentNumber"])
print("stop", client.stop_deployment(**scope, AllowRevert=True)["State"])
for delay in (0, 1, 4, 30):
    time.sleep(delay)
    got = client.get_deployment(**scope)
    listed = next(row for row in client.list_deployments(
        ApplicationId=app, EnvironmentId=env,
    )["Items"] if row["DeploymentNumber"] == scope["DeploymentNumber"])
    print(got["State"], listed["State"])

This was reproduced through actual boto3 requests, not an in-memory model test. Owned fixture resources were removed through the public APIs and the disposable container was removed; unrelated container identities were unchanged. No production AWS execution is claimed.

[AndreKurait]

Retested on 2026-10-10 against unmodified localstack/localstack-pro:dev, reporting Pro 2026.10.0.dev101 on Linux x86_64. This still reproduces.

StartDeployment returns BAKING. StopDeployment then returns ROLLED_BACK, but both GetDeployment and ListDeployments report BAKING immediately and after 1 and 5 seconds, then COMPLETE after 35 seconds. The stopped deployment does not retain its rollback state.

Resolved image index: sha256:a9b3e374db5310d706c6676a498f22b2b33a45a110ec79dbd7c41f07209c7e28; amd64 image ID: sha256:9581779d9198ddfa6f1a300760edaf7f28052eb2cfdd8f713bb528b186e0cbcc. Public SDK requests only, with no source patches or extensions. Disposable resources and the container were cleaned up afterward.