This looks reasonable.
You might want to connect the Opamp to a 3.3V supply - OPA340 is a rail-to-rail device so it will have no influence on the output during normal operation. Whenever a failure on the input occurs (e.g. input voltage is too high or ground connection is bad) the signal to the uC can never exceed the allowed range.
One thing to consider: If the source is low-impedance and the ADC is high-impedance, one could lower the value of the voltage divider, e.g. 1.7k:3.3k, add a capacitor of e.g. 100n to 1u and dismiss the amplifier. This might introduce some measurement errors, e.g. if the the ADC input has a large capacity on the input and hence the inrush current during the sampling period might lead to a voltage drop.
Why do you suspect that the output voltage could instantaneously exceed 5 V? This cannot practically happen for 2 simple reasons:
- As long as the input voltage to the divider is in the range you mentioned, the output voltage can never exceed Vin x (33/50).
- Part of the reason why voltage followers are so commonly used in signal conditioning circuits for MCUs is that they provide electric isolation.