This is false, I just ran my own thorough reproduction script on macOS with latest 1Password app and op cli. Approve a prompt once in a shell session, and all subsequent read attempts of any other secret in that vault from the same shell session (including from subprocesses) don’t result in additional prompts within the time threshold. It’s not at all granular.