A container whose change failed is billed for the size it really runs

On Platform you change a container or a job in your portal, and we carry out that change. If it fails three times, the block shows as failed, and your container keeps running on its old settings, so your site stays up. Until now we didn't measure it in those days. Now we do, as long as it runs, and at the size it really runs: if you asked for 0.5 vCPU and the change failed, you pay for the 0.25 that runs, not for the 0.5 you didn't get. If it isn't running, it costs nothing, and if your site is paused, for example at your spending cap, we stop it too and it costs nothing. If the first setup or a removal fails, you pay nothing for it at all. Every day shows as a line in your portal, as always.