Welcome to the world of sampling and signal processing.
If your device can collect samples and process them sufficiently quickly, then you can transmit a single pulse and measure the reflections. Usually that isn't the case.
In all cases the time/distance resolution is limited by the "aperture width" of the sampling circuit, dt.
Usually what happens is that you transmit a pulse and then sample the result a set time, t, later and then convert the sample to a digital signal using an ADC. After the ADC has completed, you can transmit another pulse and take another sample.
If you progressively change t to t+dt and convert the signals each change, then you will see reflections from progressively further away. Note that dt depends on the width of the sampling pulse, not the ADC conversion time. That is the mechanism by which old analogue TDRs such as the Tek 1502 can have 4GHz bandwidth yet only sample at 50kS/s, even though that looks like it violates the Nyquist theorem.
Alternatively if you take N samples at t (e.g. by adding each sample to an analogue voltage on a capacitor) before converting them, then the averaging can reduce the noise (by sqrt(N)) and increase the sensitivity/range. Of course it takes N times as long to do a complete sweep.
Less intuitively, you can also sample and convert to a digital signal N times, and then digitally adding the signals also averages and reduces the noise by sqrt(N). That can "improve" the ADCs resolution e.g. an 8-bit ADC can have 12 bits resolution - but only if the noise is random.