First a disclaimer! I'm not a sound engineer, so what I write here is mostly my own experiences or research I've done. Take it for what it is!
Read up as much as you can on recording! Use google, have a look at the article stereophonic sound on wikipedia. A very good article for me was "The stereophonic zoom" By Michael Williams. Search forums!
One of the first choices is if you're recording in mono or in stereo. It's possible to record in mono and later "stereoize" the files, but in my opinion it's more realistic and rewarding, even if it's more difficult, to record in stereo. The sources mentioned above are excellent as background information on the subject.
There are basically two schools of thought when it comes to recording organs. One is that you're in effect recording the room in which the organ is located as it contributes quite a lot to the audible sound. The other method strives to recieve as much direct sound (as the opposite to the reflected sound from the rooms surfaces) as possible to get a clear sound from the source. As we're concerned with sampling the instrument here we're likely to position the microphones much closer to the organ than one would do if one recorded a piece of music on the same instument. Or not, depending on the properties of the room…
The notion of critical distance (use google and wikipedia) when recording is interesting, but unfortunately not so easy to apply with a good result since one have to take the depth of the organ into consideration. Some suggest that the microphones should be placed really close to the sound source and moved around for optimal signal to noise ratio, but this produces problems when the recorded notes are put together in the sampleset.
In my opinion it's a balancing act. I've tried to use multiple mic positions for one organ and it definitely creates problem with the combined sound image if not done with extreme care. So, if the organ is pretty compact, I'd consider using only one mic position with a fairly good balance of direct and reverberant sound. If the organ has a "ruckpositif" then I'd consider using two mic positions but be careful to not get too different balance of direct and reverberant sound between them.
If one gets too close to the prospect, then the speaking pipes in it will be too dominant. But if you get too far away the sound will be muddied too much and lack presence. Use your ears to guide you. Remember that it's a tradeoff! You'll want to stay closer than for a normal recording. But remember to listen to stops in different divisions, including the pedal! The final balancing of strength can be done in the organ definition file as long as the overall character of the sounds work together as a whole.
Normally you'll want to get the microphones up in line with the average height of the pipes mouths. Then experiment with the distance until you find a good balance.
Decide the distance and angle between your microphones if you're recording in stereo. I really recommend the article "The stereophonic zoom" By Michael Williams! I sometimes use the "crocodile eye viewer" to judge the necessary angle for capturing the sound sources and reference the table for my microphone type to get a good setup for the microphones. If you use a measuring tape it's also possible to use some math to calculate distance and/or angle (for the true stereophonic field). See some examples on the Math example page.
Then I try to find the loudest note in the loudest stop and set the recording level after that. You don't want clipping (getting a peak at 0 dB or above) but you don't want it to be too quiet either. Try to shoot for at least around – 3 dB if possible. No compression or limiters! This is not pop music!
When recording I usually use the best quality setting that my audio interface or device allows, this means for me that 24/48 will be used with computers and 24/96 with my Microtrack.
I used to record a few stops at once (normally around four when single samples are taken) before stopping and saving. But recording one stop at a time could be considered a good practise, especially if you're taking more than one sample per pipe. Always keep note of what you record and in which order. Always record a nice and long portion of "silence" or "room noise" (with every file!) so that you can get a good noise profile to work with later.
When playing the samples I go from the lowest note to the highest one at a time. I normally shoot for at least 3 seconds of a fully developed and stable note (longer can many times be better!) and as much distance to the next as is necessary to get a quiet room in between the notes. It's generally more than you think since the microphones are more sensitive than your ears at the console. Try give it at least 2 seconds more after you've heard the note die out! The lower frequencies will benefit from (much) longer sustain sections, and the looping of the samples will also be easier especially when dealing with compound stops (mixtures and such).
In the old days sampling were very straight forward, but with recent development the added features of increased realism makes it possible to use multiple samples for every pipe, both attacks and releases. With tracker action instruments it's possible to record different attack speeds (which is meaningless in pneumatic or electric actions). With any action it's possible to record releases of differently long notes. I nowadays usually record at least one long release (if you record multiple attacks then you'll have many long releases too!), one short and one staccato. It takes long time to sample the whole instrument note by note, but it's definitely worth it! Remember to record a good portion of silence with every new take for the noise profile, especially if you move the microphones!
If you so desire it's not necessary to record every pipe to create a complete sampleset in GrandOrgue anymore. It's still the best practise to really record every pipe if you can, but it's possible to use the same note for different pitches. What I'd suggest is to have a good understanding of the pipe plantation to allow you to select wisely which notes to choose to stretch to which others in order to not ruin the stereo image too much. For instance with a C and C# layout of the pipes I can imagine that sampling 4 notes an octave would be quite good. Then you sample D to stretch to C and E, G# to stretch to F# and A#, D# to stretch to C# and F and finally sample A to stretch to G and B. In that way the stereo image will be kept fairly correctly.