This shows an exagerated case of where the total number of unique pod values and such can slow down any non-exact match label matcher at high cardinality depending on whether a prefix search is applied and also how the speed of intersecting postings lists can also affect search speed when a lot of distinct time series is matched each having a distinct label value matched by a query.
Note: As mentioned elsewhere, this is not meant to show real world performance, this is more meant to show an example of the effect of different approaches on a micro level.