1.) Camunda can be deployed in a K8 cluster https://camunda.com/blog/2019/06/camunda-bpm-on-kubernetes/
2.) www.cawemo.com or bpmn.io (last one can be used as well to build your own web modeler)
3.) Camunda offers an analytic tool called Optimize, which is part of the enterprise license. But you can use the API to get the data as well and include it in your own statistic tool. So yes it is possible to get the information
4.) There is no out of the box solution but it can be implemented. there are some examples/ solutions around. https://github.com/berndruecker/flowing-retail/tree/master/kafka or see this topic here:
Camunda-Kafka Connector
5.) Life changes of the model in runtime are not possible. Once a new model is build it needs a new deployment. Old “traffic” will stick to the old version of the model unless you migrate it towards a new version. With the EE version this is possible using Cockpit (UI for admins) and with CE via the API.
I hope that answers your questions 