We are in the process of applying an upgrade to our code generation servers.
Historically, we used a direct approach to send tasks for generation: Studio requested the codegen cluster via a load balancer. As our user base has significantly increased, more users are publishing applications at almost the same time, and we have experienced congestion issues (failed publications).
Our new approach is based on queues with priorities. We expect to significantly reduce the number of failed publications once the update is fully released.
While we are fine-tuning queue parameters, you may experience small delays (up to 10 seconds) when you send a project for publication. Once all parameters are tuned, we believe that the new codegen strategy will work faster than our current implementation.