Whether or not you need to update the model depends on how the query was done. Usually one of the first query steps is a "remove other columns" where you choose what to keep - if it was built that way, then you would need to edit that step to pull in the new columns. The query caches, so if you don't see changes you'll want to refresh the query.
By default, any columns added to a list will be available in the query editor. The default behavior is to pull all the back-end columns, so not based on a specific view, but it can be set up that way if someone went out of their way to do it.
Any time you're changing columns, you will temporarily break something, so work on a local copy first - you just have to go back and replace references. You can minimize it somewhat by setting the column names to be the same as the old ones in the model, but usually there's a thing or two you need to go back and fix...
If you're working on a shared dataset, make sure to take a good look at any other reports referencing it. If you want to test things, you can take copies of them and stick them in your personal workspace while you work out the details and make sure everything works, then publish over the old ones in the original workspace when you're done. For the child reports you can modify which dataset they're linked to, so you can swap the copies to point to the dev model in your personal workspace while you work on it, then publish over the model.
