My reason for this is I have found it difficult to find buck converters with the set of characteristics making them the optimum choice under all scenarios. Ideally one wants high efficiency, decent output current (~500mA), low differential vin/vout and low quiescent current. That needs to be combined with low cost and simple design. LDOs with LIPO batteries will have an overall average efficiency of around 87% (for 3.3V op), not as good as a good buck converter, but they can have low quiescent current and provide adequate power over the full useful range of the LIPO cell.
My main differentiating factor in making a choice between buck and LDO is whether deep sleep is used and what duty cycle is used if so. For applications where there is no deep sleep or the duty cycle is fairly high then I use buck converters as efficiency is the primary factor in determining useful battery life. For deep sleep applications particularly with low duty cycle then quiescent current becomes the major determining factor in useful battery life and at the moment that pushes me towards using LDOs. Some buck converters do have separate enable signals to reduce their quiescent current down to very low levels and that could be a way of making these work well at the expense of some complexity of control circuitry.
The other factor that keeps me using LDOs in many of my deep sleep apps is simplicity and cost combined with getting acceptable battery life anyway. I have security sensors running on 400mAh LIPOs that keep going for over a year between charges.
Running everything at lower voltages (2.8V) sounds interesting and could certainly change the choice threshold for some applications. A benefit of the 2.8V is that the deep sleep current of the esp8266 also drops quite a lot. I would also be interested if there is any data or links on the effect on rf power (range) when operating at lower voltages.