The slow-down circuit consists of the top three gates, R3, D1, C2, R4 and C3.
R2 and C1 are not needed. They do nothing. They circuit has never been tested or checked.
Incomplete, wrong, wrong, and wrong.
I know this is an old thread, but it still needs correction in case someone stumbles across it when searching for a similar circuit.
Colin's description of how the circuit functions is incorrect. R2 and C1 (and R1) are critical to its operation. Also, not only has the circuit been "tested and checked", it has been for sale around the world for many years. It works. For a discussion of how it works, here is my analysis from another forum:
That oscillator isn't what you think it is. IC1D and IC1E form a non-inverting buffer with hysteresis, which can act as a latch. It is IC1A with feedback through D1 and R2 that turns it into an oscillator. The original designer is playing some games with the resistor value ratios. For example, R1 can override the hysteresis feedback coming in through R2, and R4 might figure in there depending on the output state at pin 2.
As the overall circuit slows down, R3 always discharges C3 through D1 at the same rate. If you tied the three inverters (pins 13, 3, and 5) to pin 10 instead of the switch, then the LEDs would blink off briefly when changing from one LED to the next, and as the circuit slows down the off time would be constant and the on times would get longer and longer until the circuit stops on its final value. VERY nice.
ak