Several factors influence whether a source of gravitational waves at a certain distance is observable by a certain instrument. One way to compute the limit to the distance is explained in Abadie et al 2010 and is as follows:
Distance. The amplitude of gravitational waves decreases roughly with the inverse of the luminosity distance $\propto D^{-1}$. So sources that are further away will be more difficult to observe and at some point they will be less/not visible.
Power of the source. The stronger the source the more easy it's signal can be detected. For binary systems with larger total mass $M$ and larger reduced mass $\mu$ you will observe higher amplitude waves. The amplitude of the signal $\vert \tilde{h}(f) \vert$ can be expressed as (the expression is from Abadie et al 2010 see Finn and Thorne 2010 for a derivation)
$$ \vert \tilde{h}(f)\vert = \frac{2c}{D} \left(\frac{5 G \mu}{96 c^3} \right)^{1/2} \left( \frac{GM}{\pi^2c^3} \right)^{1/3} f^{-7/6}$$
Sensitivity. The detector can be more or less sensitive. The more sensitive the detector the lower the luminosity or distance of the objects that it can observe.
The sensitivity can be expressed by the noise power density $S_n(f)$ (as function of frequency $f$) which is specific for the instrument (and you will be able to see graphs of this in many publications). A signal can be observed if it is stronger than the noise. Abadie et al 2010 use as limit a (conservative) signal-to-noise of $\rho$ =8, which means that the signal must be 8 or more times stronger than the background noise in order to be detected.
This signal-to-noise ratio is determined by an integral of the ratio of
the frequency-domain waveform amplitude $\vert \tilde{h}(f) \vert$ and the noise power density $S_n(f)$.
$$\rho = \sqrt{4 \int_0^{f_{ISCO}} \frac{\vert \tilde{h}(f) \vert^2}{S_n(f)} \text{d}f }$$
where $f_{ISCO}$ is the frequency of the innermost stable circular orbit of the binary system
In that article (Abadie et al 2010), the limit of the distance for the detection of wave events from binary blackholes with mass $10 M_{\odot}$ was estimated at 2187 Mpc which is pretty close to the distance of 2840 Mpc estimated for GW170729 (which is heavier).
Note that the limits for binary neutron stars are more often reported and are easier to find. For instance in Moore et al 2015 you can read in more detail about the increase of the limit for LIGO from 80 to 100 Mpc in the recent years. The first image shows plots of $S_n(f)$ as function of $f$ and of $D$ as function of time (during the experiments improvements have been made and the distance was changing).
In Abbott et al 2016 a computation is performed to determine the probability to observe a particular event at a certain distance. The distance, for $40-40 M_{\odot}$, ranges up to a roughly $z=0.6$ (or using $d \approx z c / H_0 \approx 0.6 \times 3 \times 10^5 / 74.2 \approx 2.5 Gpc$), which is plotted in the last figure of that reference.
Conclusion: the observation of GW170729 at about 3Gpc is about the limit of the current instruments
References