I know typically it is not possible to sense BEMF voltage when using FOC but while playing around with some other controllers and reading data sheets I found they were using a techniques to extract phase voltage data in addition to current sense. The first one I found was on Ti instaspin, the sensing layout is identical to the VESC with one small change, phase voltage sense must have a filter capacitor (0.1uf on reference designs) and the software must know the frequency of the filter, as to what it does with the information and phase voltage, though manual loosely implies its to reject PWM signals and some sort of clarke transformation is being applied, what it is being used for is unknown as their FAST estimator is closed and secret. I have found instaspin relies heavily on whatever it is using it for, removing those vsense filter caps cause it to fail, interestingly when a VESC has them it becomes possible to see voltage sine waves under load in VESC tool.
Another I found in a dedicated motor control IC, when using sinusoidal mode it was inserting a small dead time where the estimator believed BEMF zero crossing should theoretically occur and compared the results from the inserted deadtime. This one was mainly to compensate for a very basic current sense system and may not offer any possible enhancement over multiphase current sensing.
There's very limited FOC motor control information online, I was interested to know if additional sensorless sensing enhancement were considered but not found useful in testing or if only current sense data was considered in the initial design and perhaps performance could be further increased.