Here's a thought for those who "fit an SPD regardless" ...
Do you also insist on routing the phone line, TV aerial, and any other external service via the CU and fit an SPD in those services as well ? No ? Well your SPD in the CU is going to do diddly squat against lightning induced surges
Also, the SPD isn't going to protect against a lost neutral - with the let through energy required to control the L-N voltage in that scenario, the overcurrent protection for the SPD will trip, or some internal protection in the SPD will trip/blow, or the SPD is self combust. Think about it, say it needs to pass 10A to hold the L-N voltage down to 240V, it's going to be dissipating around 2.4kW - in a tiny plastic package with limited cooling.
So we're fitting an SPD that isn't going to protect against lightning induced surges, isn't going to protect against a lost neutral and the associated over voltage. Seems like very good value for money then.
Now, a bit more on that first para ...
I do have first hand experience of what lightning can do to stuff - I've experience of sitting down with a pile of new serial port cards to fit in the PCs, stacking a pile of monitors on a pallet to go off for repairs, and going out to the bosses house to replace the fax yet again. These are the only failures I could attribute to surges. None of them would have been helped in the slightest by an SPD as required by the regs.
Surge protection, and it's related science of interference control, is a whole subject in itself and I know enough to know how much I don't know.
Lets start with the simple one - the fax (and answering machine, and cordless phone, and ...) at the bosses house. Common factor ? Connected to both mains and phone line - the services coming from different directions on overhead lines, i.e. the perfect setup for such problems. An SPD in the CU would make diddly squat difference.
Firstly, the impedance of the earthing conductor at the frequencies implied by the rise time of a lightning induced surge is high - we talk about the problems of a high impedance earth rod for a TT supply, this is a similar problem, the impedance of even a few meters of 16mm2 earthing conductor will be high enough to render the SPD ineffective. And of course, for a TN-S or TN-C-S supply, our earth comes in along the same cable as the supply so will pick up surges along with the L & N conductors.
If the surge comes in along the phone line, then the SPD in the CU will do diddly squat.
So in either case, the effect on the differential voltage these sorts of device connected between different services is minimal - and it's that which kills equipment.
You can ONLY protect against that sort of thing by routing
ALL the incoming services via one point. I really do mean ALL - that's the lecky, the phone, the TV aerial, the satellite dish cables, cable TV/internet if you have that. The all have to route via one point, so you can fit SPDs to all of them, and have those SPDs share a single common grounding point - the main thing being that they have to be tied together solidly, like being bolted to one bit of copper bar. A few feet of bonding cable will seriously reduce the effectiveness. Doing it this way, the entire installation will share a common equipotential zone - we're familiar with those and the reason for them aren't we ? - even if that equipontial zone pops up to a few kV for a few microseconds. As I said, running a bonding cable across the house to where the phone lines comes in won't help - it'll be open circuit to the sort of signals we are trying to protect against.
A surge protector socket strip is a similar proposition - but only for the equipment plugged into it, and only if all connections to the equipment pass via it. So that means the phone line for modem/internet and network cable. but just having that surge protected socket strip connected properly may "export" the problem to other equipment in the house - remember it's "earth" connection is high impedance to the problem signals.
This "thick cable is high impedance" issue has been demonstrated to me, and I do vaguely recall the theory behind it from uni and other interests in electronics. Specifically when I started work, as an apprentice we had loads of demos etc - you know the routine, how not to kill yourself (or others) on company premises, how not to burn the place down, etc. One was related to EMC (electromagnetic compatibility). In this case they had an expensive military grade filter mounted on military shock mounts to a sheet of metal. Connected through this filter was a small DC motor. Motor on, amplifier connected to supply picks up the interference for all to hear. Filter is connected to metal sheet by about a foot of thick braided cable - think the earthing strap on the car engine designed to take all the starting current. Yup, the filter is ineffective due to the impedance of this short cable. Instructor presses the filter to flex the mounts till the mounting bolts touch the metal sheet - noise disappears.
Oh yes, those serial cards and terminals. This goes back to one of my previous employers, though at the time I (with some colleagues) provided IT services to them. They had a lightning strike not very far from the site - from the description maybe 1/4 mile away across the fields. This created massive earth potential differences - even within one building. At the time nearly everything was "character based terminals connected via serial lines" to a central system - RS232 is designed for +/-25V and takes a dislike to "much higher" voltages fed into it. Each piece of kit had it's own earth potential determined by the local mains earth - with signals coming from another piece of kit with a different mains earth reference. Even stuff connected to the same DB got affected, but mostly it was case of the further away, the more likely to be damaged - and not a single serial port survived by the time the equipment was the other side of the main building. Note that in this one building, one single mains incomer, submains to DBs dotted around the building. The other building shared the same substation (we had our own in the back yard) and no serial links survived there.
Luckily this was back when serial ports weren't integrated onto motherboards, so it mostly just blew the port and we stuck a new card in. The terminals mostly just sacrificed the serial driver & receiver chips which effectively acted as fuses and saved the rest of the electronics. The central system had 64 ports in 4 blocks of 16 - 2 internal cards, one of which survived, 2 external modules, neither of which survived (but the British manufacturer amazingly repaired them FoC).
Having a bit of knowledge on the subject, excuse me if I don't share the enthusiasm for SPDs - like the freestanding socket strip versions, can be of use if used properly, but mostly are a triumph of marketing over facts. I'd also say that if you are paying £80 for an SPD, then you're being ripped off.
If you have trouble getting to sleep, or if you are genuinely interested in the physics, then I suggest a read of
BEAMA Guide to Surge Protection Devices (SPDs): selection, application and theory
It dates from 2014, so not exactly a new topic. BTW, I do disagree with some of the details - specifically where they show services entering the building from different directions where (as in the bosses house mentioned above) it's "very difficult" to get low impedance equipotential bonding across all the SPDs.