I have helped a friend with a similar issue. He was using a spinning flow sensor with magnet pickup. It worked pretty good for a while, but a couple of times the spinner got some hair-like fibers or string stuff wrapped around it. Other times it seemed like more water was flowing past it than it was sensing.
I don’t know of your situation, but since his 55 gallon drum tank full of water weighed about 550 lbs with the drum, we decided to build a platform with 4 load cells and put the drum on top. The input and output lines were flexible stainless hose. He is amazed at the accuracy of flow measured by weight over time. Weight going in during filling and weight coming out. We used a load cell board that has 4 load cell summing with a UART output of a decimal value. Noting the empty weight and full weight so alarming, drum fill, or other triggers can be dealt with at any level.
The one thing he found out right away … he was losing water all the time through a siphoning effect on a pump. He never noticed that with the flow sensor because it didn’t spin when the water moved too slowly. But the weight on the data collected showed the loss dramatically. His data collection was the imp sending the decimal value to a website every 10 seconds during a specific process and a logged data point every 60 seconds when the process was on hold. During the night when the process was not running, the data was only logged every 10 minutes. Any over-fills, or low levels triggered text messages to his phone.
This does not answer your question, but it may trigger some different ideas … I just wanted to throw in some thoughts and ideas.