The Varistors RV1 RV2 RV3 provide spike energy suppression on the 3 phase input
R141 R73 R65 provide a degree of input energy absorption under surge conditions and act as informal fuses if circuitry beyond them such as varistors or other components fail.
R140 R146 R147 and diodes D12 D13 D14 supply rectified 3 phase DC combining the 3 phases to produce smoothed DC to power U6, a small switch mode offline power supply IC. They do not play a direct part in the main power analysis circuitry.
Warning - lines that cross on diagram may or may not join: Note that the circuit is badly drawn inj the original applicationm note and suffers further when copies. The crossing lines from R141, R140 and RV1 join - this is shown by a small dot. Else where in the diagram crossing lines do not join and there is no dot. When the circuit is copied, as it has been ijn this question, the dots become even harder to see. This is extremely bad practice, but is more common than it should be. A foolproof system is to have it that lines which cross over each other never join and lines which do join end at a T. This makes them duplication and folded document proof. SEE DIAGRAM AT END
The Varistors RV1 RV2 Rv3 clamp phase voltage spikes by conducting energy to neutral when the voltage exceeds their rated voltage.
R65 R73 R141 are Yageo 100 ohm 2 Watt resistors datasheet here which limit the surge current drawn by6 the varistors under extreme fault conditions. Under extreme overload or varistor failure they would act as informal fuses. They can carry about 140 mA at rated wattage and would be required to carry far far less than that under any sensible operating conditions. They can be considered part of a belts-and-braces protection scheme. The 100 ohm value of the series resistors appears to be relatively arbitrary. It is likely that the circuit would work well enough without them present and they seem more the sign of a designer attempting to cover every eventuality.
R147 R146 R140 are surprisingly small Vishay Dale 0805 resistors datasheet here
These 3 resistors plus D12 D13 D14 are used to operate a low power offline DC to DC switch mode power supply = U6 = a Linksys LNK304. The three diodes rectify the 3 phase voltages and combine them, feel them via JP6 and R148 and then smooth the combined rectified 3 phases to DC using C35.
D17 is not a zener regulator as it may appear to be - it is a TVS surge suppressor datasheet here.
http://www.datasheetarchive.com/LNK304DG-datasheet.html#
Varistor:
The datasheet for the varistors can be found here
By following a fairly convoluted path via Mouser etc you can find that the Varistor part used is shown in that data sheet as VDRS10P510xyE on page 19.
VAC operating is 510 AC = comfortably above 3 phase line to neutral voltage voltage that it is usually exposed to.
Copy of original diagram from question showing joined and unjoined cross overs.
This is a slightly silly and simplistic analogy: -
The capacitor is like a flywheel on an engine - it smooths the rather sporadic pulses of energy associated with the movements of pistons and produces a more constant drive to the wheels of your vehicle.
Without a flywheel, the jolts and impulses from the pistons would be felt all over the vehicle and things might rattle and break or even fall-off in the road. In other words, without a flywheel, all parts of the vehicle feel the effect of the pistons.
If your chip didn't have a local decoupling capacitor, all the pulses of current it draws from the supply would "shake" the power supply voltage around and cause interference on all other circuits sharing the same power supply.
Best Answer
In addition to Autistic's correct answer, the value of C9 (X rated) is a compromise between consuming excessive AC current at 60 Hz while suppressing the noise voltage spikes by about 3 dB to 6 dB at the transformer primary if power is cut off when the 60 Hz sine wave is at or close to peak current. C9 does NOT behave like a MOV or TransZorb, but an active noise filter. When the power is cut off as mentioned above there is a burst of noise, the back EMF from the transformer. This can send substantial RF noise into local air space as well as increase arcing of the switch contacts.
C9 consumes about 150 µA of current at 60 Hz (18 mW at 120 VAC), not enough to heat up anything. If C9 were large enough to suppress most of the noise burst (say 0.33 uF), it would consume 15 milliamperes of current and 1.8 watts of reactive power. The design engineer was being careful about the power budget vs. its ability to absorb a useful amount of RF noise, likely based on oscilloscope readings of the primary voltage when the switch is turned off.
Also if C9 had a high value then it could cause a brief current spike when the power is turned ON. Keeping C9 at a low value avoids this problem as well as the extra space and cost of a large X rated capacitor.