"The problem with my live circuit is I have to create spikes myself (using a switch probably) as the 2 transformers will act as filters for the circuit"
Having another switch between the transformers and cap dropper will probably do.
Are you going for maximum LEDs protection, minimum components, or maximum efficiency.
Using a 2N3904 as a constant current generator provides quite a bit of switch on and off current protection for the LEDs, Green. Blue is the mW dissipation in the 2N3904.
It would still need some HV ~50V clamping across the 33u to survive very bad switch offs.
In my endless flashing LEDs experiments I've put 500-800mA through 5mm LEDs for a few us at a time, unfortunately I don't know how well they survive 200-300mA for a few ms at a time.
I did just that (added a switch), but the results were not spectacular

The spikes (after the cap) were very short and even though I couldnt measure current, I can't really see the LEDs being damaged by it.
My aim is a) led protection (as they are not replaceable if they burn encased in concrete) b) low parts count. (consumption is pretty low anyhow less or close to a watt)
In my test circuit I went for an X2 of .22μF (so around 22mA peak to peak with no capacitor) and a 100μF/100V smoothing capacitor I had around. For extra protection I added a 20V 0.5W zener I also had around (the 6LEDs I used are a bit lower voltage than on LTspice, all together they run at 17.6V for 15mA) and a 1watt in rush resistor of 1K

. This created a quite a minuscule ripple of around 700mV and an average of 14mA. So in normal conditions this circuit will run for ever.
After 30 minutes of use I checked the temperature of the components with the scientific method of the back of my finger (after powering off) and none was over body temperature.
So to sum it up, bigger capacitors help with inrush currents, but not with spikes from other devices during operation. For that, a zener (better 1Watt) or a MOV you used in your own circuits is a better solution.
Zeners are also quite nice for troubleshooting. If I have another failure and the LED dies first (that's the most accepted mode of failure from the people here), the zener will clamp the voltage to 20V and the capacitor won't blow.. I'm really troubled by the failure because at home I have 3-4 of these circuits running for years 24/7 no problems, and have given another 3-4 as presents in the last 2 years with no problems. Maybe my friend with the 2 dead ones lives in a particularly (electrically) noisy house

I'm really touched by your transistor idea, I remember reading about it in talking electronics and still use it from time to time when I need a specific amperage value from an LED. It was actually this type of circuit that made me understand that transistors are current regulating devices (before that I was treating them like Mosfets ie amplifiers/ switches)
http://www.talkingelectronics.com/projects/200TrCcts/200TrCcts.htmlBut I always thought of it as a current limiter solution for batteries and other voltage limited devices (like 5V power supplies). So your idea is to use a bigger X2 capacitor (like a .33μF) and then current limit the LED's with the transistor. This way you can actually get away with a smaller capacitor.
extra bonus, the transistor circuit can also be used with low voltage power supplies.
I will play with ltspice a bit and report back, but for my current construction phase (I have promised 5 of these concrete cubes lol and want two for myself too) I will go without the transistor as most of the concrete cubes are ready (and have limited space for electronics) and I want to declutter my "lab" (i'm sure you have also experienced this urge)

thanks again for all the input.