I do not think that @
struppigel missed anything and on the contrary, he understands well how such attacks were performed in the wild. This attack in the wild would not be a simple deactivation of the AV. Many AVs can recognize several other suspicious features (delivery method, using scripting, UAC bypass, code and executable for unpacking, code for payload execution, code for Defender Control execution, etc.) before the script might the chance to disable protection. From time to time it is probably possible to bypass the AV protection in this way (many AVs were disabled in the past). But this has more cons compared to pros. In the case of Defender, the Tamper protection makes it even less attractive.
Microsoft did not introduce Tamper Protection to prevent such attacks as a primary infection vector. The reason was preventing similar methods performed in the wild on already compromised systems for persistence and lateral movement. Simply, the disabled Defender protection is re-enabled by the system after some time.
Edit.
A similar misunderstanding is often related to DLL hijacking. This method is not used as a primary infection vector, too. It could be used, but this has currently more cons compared to pros. It is used in the wild on already compromised systems to make the attack more dangerous (UAC bypass, stealthy persistence, etc.).