This has actually been the standard practice by networks to call races only after they are >99% confident of its expected outcome. I can cite numerous examples of races being called well after seasoned observers had a good general sense of its expected outcome - one such example is the Montana Senate race in 2020, where the AP called it almost four hours after polls had closed, even though (thanks to a certain poster's information on key county benchmarks) I had a good general sense of which candidate was on track to win as soon as the first few counties started reporting results.
In that case, AP shouldn't have called AZ or NJ-07 when they did. Although Biden and Malinowski both did end up winning, there was a reasonable chance based on ballots counted after election day that Trump could have taken AZ or Kean could have won NJ-07.