Kubernetes application UI not robust to unexpected delays in deploying apps

On gitlab.com, I set up a Kubernetes integration manually with an Azure Kubernetes cluster.

Then I installed the Helm app using the UI, which went OK. When that said "installed". I pressed the button to install ingress. That produced a kubernetes error message about a tiller pod not being available.

I refreshed the page and tried pressing the ingress install button again. This time it gave me a different kubernetes error about "configmaps: values-content-configuration-ingress already exists". The error message persists regardless of whether I keep refreshing the page or pressing the install button.

To fix it I has to use 'kubectl delete namespaces gitlab-managed-apps' to remove Gitlab's stuff from my cluster and restart; I probably had to remove and recrete the cluster integration in the gitlab UI too.

My conclusion: the gitlab UI did not gracefully handle the fact that the Helm tiller pod was not immediately available and so the ingress installation failed.

A little bit related is https://gitlab.com/gitlab-org/gitlab-ee/issues/5645