The test statistic for McNemar's test doesn't depend on the total number of pairs (i.e. sample size) and only depends on the total number of discordant pairs and how the number of discordant pairs differs between the two groups. Based on this simple fact, one might feel there is no point to collect more pairs (N) while comparing the discordance between two groups.

Does the total number of pairs (N) really not matter for McNemar's test?