The first thing you must understand is the a.c. input to the emonTx V3.4 is also a power supply. That is not the case, and it never can be the case, with the “simple schematic” arrangement that was used before the emonTx V3 appeared.
The emonTx V3 arrangement is two very simple circuits that are superimposed on each other. If you ignore the measurement side, the power supply is a simple half-wave rectifier using D1 and C18 to provide a supply for the MCP1754 regulator. R29 is there to limit the current so as not to put too big a dent in the a.c. wave that you are trying to measure.
The measurement side is then exactly the same as any audio amplifier where the source signal is referenced to ground. The alternating voltage from the a.c. adapter is a.c. coupled via C17 and divided down by R13 & R14, with R11 & R12 with C15 providing the mid-rail bias voltage. R28 is there to limit the current in the protection diodes, and plays no part in normal operation.
There is not - at d.c., but of course there still is to a.c., via C15. Yes, really: C15 puts the ‘bottom’ end of the divider at an a.c. ground, the same as the sleeve of the a.c. input connector - and that’s an a.c. ground as well as a true ground.
If you carefully check the voltages, there is always a positive voltage applied to it, so there is no need. And of course, a 1 µF non-polarised capacitor would be huge, and would be expensive to fit, probably needing to be done manually.