But when it matters, it matters — there's a huge productivity difference between waiting 5 minutes and 30 minutes when making staging deploys to iterate on bugs (and sometimes I'll do this 5-10 times in a day).
My way around this that's less than $500/mo and indefinite wait times is to have an EC2 image running as a jump box. Then anything downloaded is already on AWS's network, making ECR uploads much faster.
But when it matters, it matters — there's a huge productivity difference between waiting 5 minutes and 30 minutes when making staging deploys to iterate on bugs (and sometimes I'll do this 5-10 times in a day).