The melting case is not the result of wrong fuse (even correct fuse won't help here), but of the too high output power, above what the design could thermally manage. For this you need really good control over the maximum power.

The cases, where the charger delivers lower than rated current I won't name "bad" or "dangerous" (only when speaking about the load characteristic, not the other safety rules). I would more consider them as well behaving for the purpose of a charger:
- You do not need any accurate voltage. Just anything between 4.5..6V is just OK.
- If the voltage start to drop at the maximum current, it usually mean shorter charging time compare to the chargers maintaining constant voltage to higher current levels.
The reason stay in the way, how the charging controllers within most mobile applications are designed: They are a linear regulator with a current limit, but in a very tiny package. So in order to prevent them from overheating, they reduce the current limit as they heat up. So consequently when the battery have still low voltage, the charger controller chips reach their maximum temperature and reduce the current, usually way below the nominal ("0.5A" charger tend to drop down to 200mA)
So when the input source (= the "USB") drop it's voltage at currents below that limit (so e.g. at 0.45A for "0.5A" device), the voltage drop across the charging controller remain low, so does the power dissipation, so the charger controller remains fully ON and charges the battery by full 0.45A (instead of the 0.3A on the "hard" 5V USB supply).
And the voltage regulation does not have to be so excellent - the only purpose of these sources is to charge the battery, so no need to extra filtration or accurate voltage. It only should be sufficient to charge the LiIon and keep some headroom and still be able to deliver reasonable current (so 4.2+0.2V=4.4V at 0.45A is enough for vast majority of devices), while stay below the maximum operation range of the charger controllers, so below 6V (open circuit), that it is.
So two transistors and few passives in a selfoscillating circuit do the job very well, only the mechanical construction should be sound for a mains powered class II device...

Have a look at some of the chargers there for whats so bad. Examples ?

 - The melted charger - if it can melt like that without hard failing, it can probably heat long enough to start a fire. A single "bang" failure is not nearly as dangerous

 - Clearance between input and "S"ELV is as thin as 0.5mm

 - Fusing - just look what there is, 0.125w resistor with <0.5mm between its soldering pads on the bottom, just some snaky PCB track with 0.3nn overal between ends

 - Ripple on the output can scew up some electricis pretty bad

Have a look at some of the chargers there for whats so bad. Examples ?

 - The melted charger - if it can melt like that without hard failing, it can probably heat long enough to start a fire. A single "bang" failure is not nearly as dangerous

Exactyly. And here the fuse win't help - even a 200mA fuse allow 40W of dissipation, what is way too much for such small box...

- Clearance between input and "S"ELV is as thin as 0.5mm

 - Fusing - just look what there is, 0.125w resistor with <0.5mm between its soldering pads on the bottom, just some snaky PCB track with 0.3nn overal between ends

Well, these belong to the "other issues" and what I meant by "not being sound" design. And that is the part, what make many of them quite a hazard - and basically disqualifies them, regardless how good is their functional output characteristics. The safety should simply not be compromised - it does not cost as much...

- Ripple on the output can scew up some electricis pretty bad

Once it does not exceed the maximum operating voltage, there is of no problem at all. For sure, it does cause a disturbance to an audio and/or video signals, but here we are talking about charging a battery, so when it does not overstress anything, there is no problem at all. And as the phones operate at at least 900MHz and their RF part is quite well "shielded" behind at least two regulators, there is no noticeable disturbance on these frequencies either...

And don't forget there are quite good quality (ceramic) capacitors of few uF on the input, what form quite a good filter together with the cable impedance. The measurements were done with a resistive load, what is quite high impedance (it's resistance) for the HF, but the capacitors in the cell phones are way lower impedance on the ~100kHz these chargers usually operate.

And who said there are no few small orange electrolyte cubes in the phone itself ? Those definitely wont like the ripple (the HF one not the 50-100 hz)

And who said there are no few small orange electrolyte cubes in the phone itself ? Those definitely wont like the ripple (the HF one not the 50-100 hz)

50/60Hz ripple current is way more stressing on the capacitors than the high frequency. But that would mean way higher ripple voltage as well, so it is of no concern on the secondary side (where the converter's regulation loop suppress nearly all the low frequencies)
And for the high frequency ripple current the small electrolytics form too high impedance, so the current goes rather to the ceramic ones.

And don't forget the phone itself draw amp peaks during transmitter activity (up to 2W peak output, what mean ~2..3A peak from the supply rails) and in the smart phones due to the processor activity as well.
So the charger, even when with high ripple, does not mean as much load as the phone itself...

All interesting reading , but again, (and the same for CFL), it seems to be a problem for you guys in the US, and your 120volt supply?
I may be wrong, but I have never come accross, or heard of a fire hazard with our 240volt versions, (I think Medved has explained this with 120volt CFL before on here, but I don`t remember where it was at the minute)!

I`m new to using retrofit LED, but all the 240 CFL I have used so far just `go out` at EOL?

With LED there is not the voltage doubler problem anymore, as the LED ballast could be designed to use virtually any input voltage with the same topology, so all the fundamental problems are exactly the same.
The fact you haven't come in contact with that problem does not mean the problem is not there.
Don't forgot, than the US is 240? million people sharing the same language, so a single such case within these 240 million people get the same attention as a single case in e.g. a 10 million Austria. So only from that there is up to 25x higher chance in the US for such thing becoming widely known.