Job
Just a new (or known to be good) coil is probably your solution as the starter isn't reducing the voltage at the coil as it may with a larger engine.

Another test if you wish, would be to carry your spare (dodgy!) coil in the car, and as soon as you have a starting problem, swap the leads and wires to the spare (cold) coil and lay it some where suitable. Try starting it. if it starts, that confirms a dodgy coil when it gets warm.

I've heard rumour of them being cable tied together for a quick swap.

Roger V. If I strap a second coil to the suspect coil, won't it be hot like the one that is suspect, or do coils get hot when in use irrespective of the temperature under the bonnet. Earlier on in this discussion, I had changed the coil for an older one (Which of course could be suspect as well).
Has Hot Start always been a 2CV problem, or is it something to do with modern replacement bits ?

Age of the coil. A friend has very difficult hot start problem on his -57 DS. Many years tried to find the reason. Usually it worked well, and if started after very short time again, there wasn't any issue, but over five minutes or so it was possible to get engine running only after over half hour...

Once then we tried if there is any kind of spark when it didn't start, we find there wasn't. But short after the hood was open, spark was there... And the problem was solved by new coils...

And how this is related to 2cv? Well, on those early DS, there is two 2CV coils and two points on "distributor" (what is under intake manifold, very hard to maintain)...

next daft question, what goes wrong inside a coil? I would have thought that if the insulation broke down enough for it to short out the windings, then the coil would fail just as much when hot or cold. Lets assume there are 5,000 turns on the secondary (Probably many more). If say 40 turns shorted out, it would only alter the O/p voltage by a few volts. If the primary shorted out 40 turns, then the O/p voltage would probably drop by a few thousand volts, but if the primary shorted out, I would think the coil would die as a 12volt circuit wouldn't jump the gap between the shorted windings.

Does it matter which way you connect the points/ignition wires on a coil ?
If there is a "wrong way" would this lead to coil failure ??

Original Citroen coils are filled with oil to help keep things cool. The oil dries out over time and causes the issues. I've converted mine to one of the new type sold by ECAS. https://www.ecas2cvparts.co.uk/ignition ... 011f07d324

EDIT:-The new resin filled coils MUST be fitted correctly, but the older oil filled versions can cope with swapping around the two LT wires. From what I've been told, swapping the two wires increases the service life of the spark plugs (but don't quote me on any of that).

Jonathan,
I doubt that it's 'oil drying out' that's the problem. In my experience, it's overheating which will damage a coil, possibly by oxidising the insulating oil, as can be seen in the photograph below.
Accidentally leaving the ignition switched on without the engine running is the quickest way for that to happen.
.
I've got resin-filled coils on two of our Dyanes and regularly reverse the polarity of the low-tension connections, just as I've done previously with oil-filled coils, in order to equalise the wear on the plug electrodes.

There's never been any problem with starting and both cars will return well over 40mpg on a decent run out, so I'm not convinced by the advice that coil polarity is critical for these resin-encapsulated coils.

Primary winding of a 'cooked' ignition coil by slcchassis, on Flickr

The coil doens't look difficult to redo your selves? Is the thing filled with PCB oil? So yes than be aware it's toxic.

@AZL57
Like the idea, however the secondary coil is lots? of very fine gauge wire, but is it the primary coil that fails??

Those what I have tested, have all broken secondary coil. I'm not saying the primary won't fail, but what I have experienced about them they usually fail secondary coil...

Jasu,
could that be because the primary circuit only needs to cope with a peak of 'about' 300 volts, whereas the secondary can be 30,000 volts, depending when the owner last checked the plug gaps...