Yes, I mean something as simple as a scatter plot, or histogram (which may or may not reveal normality), in the case of discussing outliers you want to understand that visually I think. Again the outlier detection approach you are trying is based on boxplot , so boxplot is a reasonable way to try visualise also.
I'll add some context. In my field finance and credit risk. If I'm building models, one of the biggest challenges can be data quality. This means that every so often, someone has typed an absurd value into a field on a terminal corresponding to something. It's reasonable me to try and detect these and reject them as outliers because I doubt so strongly they reflect the truth of the account, customer, transaction or whatever. The model should be built on as much data as possible, or rather quality data. If a magical fairy could guarantee that all the data you received perfectly represents the reality of what you are analysing, would you have a motive to remove outliers per se? I mean, you might depending on what you are doing. But for most purposes you probably wouldn't ? Also if I told you the machine used to make your measurements can be expected to give a spurious result an order of magnitude out on every thousand samples, it's reasonable to think that that information would guide and justify an outlier detection and removal strategy.
But... If you are modelling the total system, subject under study and the machinery as a totality maybe you wouldn't even exclude them, because if you wanted to predict future performance , and there's no plan to improve the machines maybe the predictions should sometimes predict these odd tails. Am I making sense or just one long, out of my depth, ramble ? Anyone ?