Experiences on differential GPS?

Has anyone ever tried to do differential GPS and has some data on it?

In principle, it seems straightforward: Accuracy is limited by atmospheric errors and close-by GPS receivers will experience similar errors*. So subtract the error measured by a stationary receiver with a known position from the data of one with an unknown position to get a much more accurate position.

Has anyone tried this and can give some estimations on how the error signal will look? How accurate will this method be when the two receivers are close by, say <1km? How often does the error on the stationary receiver have to be sampled, and how accurately does it have to be synchronized?

What I currently plan is two MT3333 receivers equipped with RFM95 lora transceivers and BME280 barometric pressure sensors for better z-accuracy.
The LoRa radio is just for timebase synchronization right now (exchange one message, than count GPS ticks to stay synchronized), as accuracy in post-processing is sufficient, but I wonder what i could achieve in life correction with such a system.