pulumi-awsx: Fargate service updating failing with timeout

I am using the following code to deploy container (some things with certificate, route53 dropped out to simplify )

 const nlb = new awsx.elasticloadbalancingv2.NetworkLoadBalancer("some-nlb", {external: true });
 const tg = nlb.createTargetGroup("some-tg", { port: 3000 });
 const listener = tg.createListener("some-http-listener", { port: 80 });
const listener2 = tg.createListener('https-listener', {
    certificateArn: certificate.arn,
     loadBalancerArn: nlb.arn,
    port: 443,
    protocol: "TLS",
    sslPolicy: "ELBSecurityPolicy-2016-08"
}); 

let service = new awsx.ecs.FargateService("some-app", {
    desiredCount: 1,
    taskDefinitionArgs: {
        containers: {
            myapp: {
                image: awsx.ecs.Image.fromPath("app", "../"),
                memory: 512,
                portMappings: [listener],
            },
        },
    },
});

It deploys correctly, but if I try to run it again, it stucks on the updating taskDefinition, and throw timeout error:

    ├─ awsx:x:ecs:FargateTaskDefinition  
+-  │  └─ aws:ecs:TaskDefinition         replaced                [diff: ~containerDefinitions]
    └─ awsx:x:ecs:FargateService         
~      └─ aws:ecs:Service                **updating failed**     [diff: ~taskDefinition]; 1 error


error: Plan apply failed: 1 error occurred:
       * updating urn:pulumi:dev::drjones::awsx:x:ecs:FargateService$aws:ecs/service:Service::service: timeout while waiting for state to become 'true' (last state: 'false', timeout: 10m0s)

And it always crashes on process updating taskDefinition, after this, I have +1 running task, but not replacing existing. But my changes are successfully applied (which I made in app code). So I see the changes which I made in container, but script fail and I have 2 tasks not one.

After some time this second task is stopped on aws. But script on the next pulumi up run tries to delete task from previous deployment.

Weird situation, which affects consistency, and makes it impossible to do updating correctly. Anyway it looks like some processing up to 10 min and then fail.