Protection from AC transient for Power meter

This can be a pretty complicated topic.  Since you're measuring 40VAC I assume this is not any kind of legally regulated utility metering situation with strict accuracy requirements, and maybe not even a very high volume application with strict cost / size / parts count limits.

In general these are the protection options available for any situation:

A: Fuse.  Protects from gross over current conditions.  It doesn't typically prevent damage to the electronics components or even the circuit board.  Typically the fuse is provided in a power application to protect the wiring and assembly itself from something getting so damaged / hot from over current that there becomes a fire hazard or situation where the insulation on the circuitry can start to melt.  Some smaller fuses are of course meant to protect against less gross levels of faults, but in general the equipment is often considered to need service if the fuse blows, and it wouldn't be a guarantee that nothing else (delicate ICs) was damaged in the event that blew the fuse.  Also one would have to ask WHY the fuse blew, since that is often indicative of some internal or external condition that may need to be repaired before putting the equipment back into service.  So if you have inputs that are supposed to be low current circuits, fusing them with low blow current fuses might help protect parts of the equipment or installation.  If this is a power application where the meter is metering the entire current flow going through to some other kinds of equipment from a power supply then it MAY also be reasonable to insert additional fuses at the input to your meter where it gets its power from the power supply, though if you control the environment you may also have some degree of reliance that the power source itself is going to be fused at some certain level that you can count on, making redundant (sort of) additional fusing at the meter power ingress.  Sometimes there are reglatory / engineering reasons why having multiple fuses in series on a given power branch circuit is not recommended, but if you're controlling the specification for the meter, it may not hurt to add some fuses to protect against gross faults as a "last resort".  Keep in mind the interrupting voltage and interrupting current ratings of fuses.  If you have fault current levels too high or if the voltage is too high even a "blown" fuse may not stop the current.  Also DC vs AC and frequency matters.  And of course any needed type approvals for the fuses applications.

B: Circuit breaker .. another kind of over current interrupting device.  The same considerations apply as with a fuse.  The benefit is that they're easier for someone in the field to reset if they accidentally are tripped due to an external situation which has caused no permanent harm to the equipment or the installation.  Accidentally plugging the circuit into the wrong polarity or voltage line during test or installation would be a typical "oops" case for a fuse / breaker blowing which might be harmless to the equipment if the protection measures can fully deal with the fault situation until such time as the fuse / breaker disconnects or the problem is resolved manually.

C: Crowbar.  This is something like a MOV, Zener diode, gas discharge tube, sidactor, diode shunt to the supply rails, et. al.  During a fault they shunt fault current away from more sensitive parts of the equipment by conducting heavily through a safer return path like to ground or back to the power line etc.  Some of these devices have especially limited lifetimes like MOVs, and some like MOVs, Zeners, Sidactors will tend to be more dissipative of power while they're conducting than others.  Some like diodes or discharge tubes or carbon piles or whatever can conduct without all that much dissipation of power and without substantial harm to the device's future integrity as long as their ratings are not exceeded.  With any surge suppressor / absorber or crowbar, though, there will be limits of voltage, current, thermal dissipation during a fault, energy handling capability during a brief or extended fault, etc. that will dictate how much fault voltage / current / energy can be handled safely before the device is compromised and they're unable to safely handle more.  Typically you'd design the equipment so some kind of fuse / breaker would permanently disconnect the circuit and interrupt the fault if the fault current / energy was above a level that the wiring or crowbar / surge suppression type of components could safely handle. 

D: Snubber, surge suppressor, sidactor, MOV, gas discharge tube, etc.  These typically work like crowbars above, with the added condition that they're energy filtering / absorbtion devices that would tend to dissipate more energy in a fault condition than some other kinds of protections.  They're useless just like crowbars for continuous / long duration faults in general, and you'd use a fuse / breaker in addition to interrupt high energy / long duration faults.

E: Transient protectors .. these may be MOV, sidactor, snubber, GDT, etc. but geared more for protecting against things like EMI, lightning strikes, ESD surges, etc.  The frequency spectrum, voltage, current, energy of these types of events requires some kinds of different planning / protection than low voltage long duration events like a short circuit or accidental crossed wiring error or accidental polarity reversal etc.

F: Polyswitch or similar.  These are like fuses but they are "self healing" such that they will become much less conductive during a slow over current fault condition, but will cool down and become "normally" conductive if the current drops back to nominal levels due to a fault condition clearing.  They have not insignificantly small resistance and energy dissipation even in normal conducting state, so they'd be less likely to be used in series with a significantly powerful circuit of a power supply in normal operation as they'd cause some energy loss, voltage drop, and would have some lifetime and other concerns.  However if you're dealing with signal protection or low currents and low voltages, say under an ampere, maybe they'd provide some kind of useful protection in some cases.

G: the other option of course is to use some kind of active monitoring and have the circuitry itself actively protected against excessive temperature operations for key components, excessive current flows, excessive voltages, etc.  Most power supplies are geared to have some kind of auto shut down via mechanically switched relays or transistor/mosfet based switching or under other kinds of control when they detect some kind of fault.  It is useless for something fast like a ESD surge or a lightning strike in that they will not be able to react before some possible damage is done, but sometimes they can detect the event and signal a service problem and can possibly place themselves in some kind of diagnostic / lockdown state after they detect a fault until the fault is cleared automatically or manually.

So for a meter, typically you'd have some ingress over current protection ahead of the meter -- breaker / fuse.  You'd have some kind of crowbar protector between hot and neutral or power and return for each power circuit to protect against over voltages differentially on those individual circuits.  You'd have some kind of common mode fault protection like a sidactor or zener or whatever between each incoming signal / power / return line and your local chassis earth ground if your equipment is compatible and designed with such configurations.  This will protect against common mode fault voltages relative to earth potential.  It might be reasonable to have distinct series connected overcurrent protections on each egress signal / return / power line going out to downstream equipment too, depending on your design.

Furthermore you'd avoid things like MOVs just due to reliability issues if you wanted something that had a certain level of repetitive gross surge handling characteristics.  There are other kinds of protectors that would typically handle repetitive surges better, though it depends on what ratings / reliability / cost you need and can tolerate.

For metering you need to measure both current and voltage.  The voltage measurement circuits will typically have a high impedance like a voltmeter, so as long as you stay within some limits of impedance relative to the accuracy the meter needs, you can put series resistors, crowbar diodes, even RC snubbers and LC filters in the circuit with the voltages to be measured to protect the meter voltage measurement input itself.   Since there will never be much current into a voltmeter you could contemplate series connected polyswitch devices before your crowbar diodes at the volt meter's input.

To measure AC current you might use a differential voltage across a low voltage shunt resistor which is in series with the power circuit.  In this case you could potentially use isolation techniques to isolate or "float" the current sensing and telemetry transmitting components / signal relative to other parts of your circuitry.  Having it isolated and floating would make it pretty immune to certain kinds of fault conditions and of course would simplify any ground reference compatibility issues as well. In any case since you're measuring a low voltage differential signal the surge protection is simple because you never expect much voltage across the current shunt resistor and you just have to protect against ESD, EMI, lightning, and common mode high voltage faults mostly.

A simpler way to measure AC currents might be to use a current transformer (CT) where you're just measuring the secondary voltage developed across a transformer where the load is flowing through the primary.  The transformer can provide extremely good galvanic isolation between the primary and the secondary, and there is a limit to the amount of energy in transients or in sustained conditions that the primary can couple to the secondary due to transformer saturation in many cases.  So just the usual clamp / crowbar / snubber / fuse types of protections in addition to using an appropriately rated CT is all you really need.

So basically to achieve defense in depth you need some protections for gross faults like fuses / breakers, then protections against faults that are just small enough to NOT trip the fuses / breakers (fuse/breaker / other protection component failure is a possibility that merits consideration but is difficult to deal with in cases short of having fully redundant protection networks -- not a bad idea in many cases).  Then you must protect against fast transients like lightning EMP / ESD and momentary shorts. 

Add series LC/RC filtering to any inputs that can tolerate it for fast transient / surge filtering leading into more delicate components.

Add shunt/crowbar components to conduct away over voltages / bad polarities to safer current return paths away from delicate ICs.

Add sidactors or similar surge suppression components after the fuses/breakers to help control gross overvoltages temporarily until the pulse clears or until the interruption fuse/breaker trips.

Add series resistances and voltage dividers where possible before going into high impedance IC inputs so that the magnitude of any fault current or fault voltage is lessened if possible.

Use isolated transducers or transmitters or sensors if possible to help isolate different parts of the circuit so that a fault on one circuit doesn't surge through the rest of the equipment to affect other circuits and so that you can isolate to some extent the locations and types of protections you need for a particular part of the circuit.

Having a 120VAC or 240VAC accidental line cross onto a 40VAC circuit isn't THAT huge of a difference in fault voltage (x6 for 240:40) that you can't find plenty of good AC line rated protective devices and techniques that would do a fine job protecting your 40vac circuit.  It would be harder if you were trying to protect against even higher voltage line crosses, or if your equipment normally had to deal with very high power currents but I'm assuming you may well just "normally" deal with currents less than the 50A or 20A levels that common AC circuits routinely operate at on the 40VAC lines, so at least there are lots of components that are rated for such power/current/voltage/energy that can help protect your meter.  Just design the primary protection as if you'd EXPECT a 100vac or 240vac or whatever fault condition, and have that cause no worse damage than a blown fuse/breaker, and make the rest of the interior circuit survive that level of fault for either indefinitely or at least for a couple of seconds while the interruptors trigger / blow and it will be pretty robustly protected.  Add in some pulse protection for ESD/EMI/indirect lightning and AC line coupled spikes and you're all set.  Sort of a 3-4 layer defense.