Hi,
Quite a fast-moving thread so I stayed out until now, but here is my "advice".
Firstly, yes 33k is a "good" value for the serial input pull-down resistor. It's basically the same as 22k + 10k which is the pull-down provided by the "standard" programming circuit when nothing is actually plugged in. If I'm developing a "custom" circuit, I usually try to get the most "bang (not literally) for my buck", which may avoid the need to design one for another purpose later. Maybe I carry it to an extreme, but:
Personally, I would always include a "Programming Interface", not necessarily the jack socket, but just a 3-pin header (or even just the pads), rather like the "Legacy" programming socket still documented in the original Manual 1. Consider it as a "Test/Debugging Interface", if you really believe (IMHO wrongly) that you will never want to program the chip again. It will also make available a spare "ground" connection (which you no longer have on the LED headers) and the ability to connect an input switch or data, to be read by the Programming input (the only time when a DISCONNECT is required)
It's not even strictly necessary to fit both programming resistors: With a 33k pull-down (or preferably higher) on the pin, you can use an external current-limiting resistor (10k - 22k) if you decide that you do want to program/debug the chip in situ. Then, you could make a tiny "programming adapter" on stripboard that carries just a standard 3.5mm jack socket, a series resistor and a "Molex" (header type) plug.
The pull-down on Leg 4 is not absolutely necessary, particularly if no external connection is being made to the pin, but we can be a little more clever: The "unused" inputs of the Darlington chip actually have pull-down resistors of about 12k to ground. So just connect one of those inputs to Leg 4 to act as a pull-down (personally I'd add a pin/pad as well just in case I wanted an extra input later). There's also the possibility that you can enable/disable the PICaxe's internal "Weak pullup Resistor" on the "Input Only" pin (and even the programming input) to turn on the Darlington transistor, as if it were another output (or two) !
If you don't want to add pins for those "unused" Darlington Outputs, they could be connected to an adjacent output as a "Wired Or" gate. That gives the possibility of "modulating" two outputs, probably one PWM, to dim and flash the LED, or particularly to make interesting sounds.
As for the decoupling/bulk capacitor; one of the main considerations is the parasitic (unwanted) inductance between the power supply and the PICaxe. The effects are worse if the cables are longer, or the current and/or the frequency are higher, so all need to be considered. One of the problems is that the "old" PICaxes (e.g. M or M1) were slower than the M2/X2 chips, so the Original Manuals and "Legacy" boards may not include the capacitors that are now recommended or even essential. In practice, I usually just choose the most convenient "can" size, which will often have an "unnecessarily" high voltage rating, but that does no harm. However, choose the voltage rating immediately above your (maximum) supply voltage, if you think a higher bulk capacitance may be needed.
Cheers, Alan.