Believe or not, but you can actually scan your MIDI setup for both the EX-800 and the Poly-800MkII. If you have more than one synthesizer of the same type, they must not be connected to the same MIDI wire, as there would no way to control them seperately. Sound data is imported automatically when you let SD request the devices' memory.
The orginal Poly-800 plays dead with respect to sysex, unless you've installed and activated an EX800 ROM (you can find the necessary ROM images and modifications at several places on the net). If you've got a modified Poly-800, it should be supported by the EX-800 adaptation.
First of all: make a backup of the data in your synth, either via the tape interface or the Generic SysEx adaptation in SD! I have no Poly-800MKII to test the adaptation with, so I can't tell if everything works correctly there. Don't rely on Program Write being set to Disable: this switch is unfortunately ignored when sysex is received.
The MIDI implementation in both machines is very rudimentary, so you'll have to live with some idiosyncrasies. There is only one dump that contains the Global data, the Sequencer data and all Sound data. Global and Sequencer data are handled with one bank and Sound Data with another. Both banks have one editor. Even though the data in both banks is contained in the same dump, you have to request them seperately. Another drawback of having just one dump message for several data types is that SD cannot import MIDI and SysEx from files, as it can't determine which of the two targets to use. If you need this function, you temporarily change the MIDI string, for instance by inserting a bogus manufacturer, for the bank driver that you don't need for the import (most likely the Sequencer Data).
Editing works as usual, an edit buffer is provided for both banks. These edit buffers exists only in SD, not in the synth itself. When you have actually changed something and close the editor, you will be asked if you want to store the data. If you answer yes, Sequencer data will be immediately written back to the synth. For Sound data, you will be asked if you want to transfer the complete bank of sounds, as there is no way to transfer a single sound. If you do this, the Sequencer data in the synth will be overwritten with bogus data (sorry, there is no way around that). Among the things that get overwritten is the MIDI channel (reset to 1) and Bend Range (almost certainly set to an illegal value). To re-install the Sequencer data you have to transmit the Sequencer bank by hand each time the Sound bank has been sent to the machine.
The Sequencer Editor adds some functionality over what the synthesizer itself allows you to do. Normally, you can only play notes from C1 to C5 and anything outside this range is mapped to the outermost valid octave (C1 or C5). The sequence editor allows you to use two more octaves at either end, however depending on what the octave settings are on DCO1 and DCO2 these notes may not play back correctly. Just experiment a bit, the extended range is sometimes quite useful. Each step in the sequence can have four (double mode) or eight (whole mode) notes, the last note in each step needs to have the step button active. If you input more notes into a step, they just don't play as far as I can tell. There is an extra text field to enter the end of sequence (Stop) as well as rests and ties. These show up in the note field as C-2 C#-2 and D-2, and you must not activate the step button for these special steps. Everything after the first Stop is ignored.
Setting the first note to Stop while the sequencer is running often crashes the synthesizer, so you may want to avoid to do that. You have to remember that due to the quirky sysex implementation this also happens when you send an edited sound to the synthesizer, so you should as well not do that while the sequencer is running. Besides, you would not be able to hear the changes anyway: while sending the sounds to the synthesizer causes them to be changed in memory, the selected sound is still in buffer and you need to re-select the sound for any edits to become audible. So if you think you don't hear some of your edits, re-select the current sound and see if that helps.
You can convert Sounds or even a complete Sound Bank between EX-800 and Poly-800MkII. Conversion to the EX-800 will strip off the data for the Digital Delay effect unit and the EQ. Conversion to the Poly-800MkII will leave pre-existing Digital Delay and EQ data alone - if you don't want the DDL on the converted sounds, just initialize the entries before pasting the EX-800 Sounds to them.