Cancelling a pulumi run with ecs definitions takes ages, now running for almost 10 minutes. Is there a way to make this quicker?

The cancellation support for the AWS provider currently waits for any in-progress the resource operation to succeed or fail, since we can't be sure the state of the resource until it completes. For ECS definitions, if you are using waitForSteadyState it could take up to 10 minutes to confirm whether the update has succeeded or failed. You could be able to ctrl-c twice to hard cancel (I believe you should get a message noting this option on the first ctrl-c).

Can confirm: I Ctrl C twice to get a hard cancel, but the operation does go on in the background, pulumi just doesn't wait for it in your shell window/terminal. If you need to be sure of the resource state you just have to wait it out.