I’m thinking about moving to Render from Heroku. The reason is the unpredictable performance of all the shared Dyno types, including the “1x” and “2x” sizes. This has been a widely reported experience.
In contrast, on Linode, performance of their shared-server VMs is rock solid. But it’s not a PaaS.
Our underlying approach to running services differs to Heroku and whilst you are able to exceed your available memory resources (R14/R15) running on Heroku, on Render, if you exceed your service’s max memory for your plan your service will be restarted. When it comes to CPU you can only consume the CPU you have available to your plan and whilst multiple services on the same node could be utilising all their CPU this would be down to us as a service provider to manage to ensure the node’s CPU is not exhausted. This is particularly relevant as you move from our shared CPU plans to dedicated CPU.