There are a number of users who will accidentally search a wild card against a large dataset from within Kibana. This can sometimes cause the cluster to freeze up while providing the results to that query. It is also possible that end users will pile on more queries on top of the initial request, only making the load on the cluster worse.
Utilizing the new ability to associate an ID with tasks (elastic/elasticsearch#27764) and the task management API, we should be able to automatically cancel a running query if it's modified or removed.
Related to #14544, #17265, #17845, #17846, #24171