Ok so the problem with filters (pinned or not) is that they aggregate, so if I want to see tags X,Y, and Z (filtered in that order) then the existence of Z depends on the existence of X and Y. I'm using tags to tag specific log lines and these are separate things (if X occurs then it is not Y or Z). So If I use filters I will actually only see the data from X since the aggregation of Y and Z produces nothing. Does that make sense?
I just looked at saved searches, that is close except you can't apply them to dashboards or visualizations only to the discover page. Which brings me to my next part about dashboards.
I think dashboards should be able to be designed so that an uneducated person can log on and tell what's going on at a high level. Now with some manual work I can, say add an "important" tag to X,Y, and Z and then filter that for display but that really isn't the best because if they want to drill down further around events X, and Y in a time frame they can't. The best dashboard would allow a person to flexibly view data like this. I do agree that you need some inherent knowledge but I think the great thing about tools (like the Elastic stack which is amazing) is that they can greatly reduce the amount of inherent knowledge needed. Imagine what I'm proposing and perhaps it will make sense:
An "Add a query" UI with the same design as the "Add a filter+" UI that also allows boolean combination of queries. Once the second one is clicked it would force you to choose AND,OR, or any other valid option (there's some room for design here but you get the point). Essentially it's just a UI for creating the query that with more knowledge someone can do in the search bar. Except the other major benefit is that it can be saved to the visualization/dashboard so that it shows by default and it can easily be changed using the intuitive mouseover toolbar (edit, delete, etc.) that the filter UI has.