The accessible prompter tests that check the none-echo mode are still a bit flaky. For example, I observed this failure:
https://github.com/cli/cli/actions/runs/14992835636/job/42120038531?pr=10975#step:5:44

For now, it makes sense to increase beforePasswordSendTimeout from 20 us to 100us to make sure there's enough time for huh to set the none-echo mode. Since the sleep is very small this seems like a viable solution. If we again see the flaky behaviour, then we might want to try another approach (e.g. waiting for the terminal configuration to be set before sending in the password/token, as suggested by @williammartin).