Nice initial project with SR04. What I ca say is that readings fluctuate and a sort of averaging is needed. Then calibration: the readings must coincide with reality on a wide range. Many projects with SR04 exist even on Hackster, so there is room for improvements.

About the sensors issue raised by "infusion": every sensor needs calibration. There are many tutorials about calibration. This particular one is easy to calibrate since it provides data at large time intervals. Take tow limit states: open circuit and short-circuit and calibrate all sensors. The readings are different, but using the "map" function in Arduino, you bring them in the same interval.

Surely corrosion exists! It is a matter of time. If a 0.65$ sensor can last 2-3 years, it is perfect.
I will post images with mine after a year. So far no sign of corrosion. Other problems can still exist: the moisture uniformity in the pot and around the sensor. A thin layer of mud tends to stick to the sensor, perturbing the readings. Maybe a mixture of inputs: humidity and time lapse are the best choice for an Arduino project.