Rowid isn’t really a computed column, so the data in that column is always there, just like name would be. It’s pretty much a basic column.
I would say however, that there could be a slight delay with other computed columns, such as a template, lookup, math, single value, etc. Those columns are computed directly on the user’s device when the data is first cached. Once it’s loaded initially, then I would say that computed column based relations are just as fast as basic columns. That is, until the app is closed and reopened again. So the answer kind of depends. I’d say in the case of your poll, both are correct and the fastest, but that’s really comparing basic column to basic column instead of comparing to a true computed column.
I have an app with relations based on computed columns and the sheet it relates to has 4000 rows. That sheet also has a lot of calculations that it performs. It’s really slow to open that part of the app initially because I believe it has to first perform all calculations in the sheet before it can completely link up the relation.
Also be careful with using unique. You are probably fine with what you are doing, but if a row in the original sheet is deleted, it can reshuffle your UNIQUE sheet, which I have seen cause issues in the past. Also, any new name would cause a much more substantial delay as your app would have to wait for the data to get from the app to the glide servers, to the google sheet, recalc the unique formula, then sync back to glide, and finally back to the app…as compared to having everything calculate directly inside of the app. If your unique sheet is doing nothing more than holding a list of unique names, then I would instead recommend the following: