Has anyone done any performance measures with this in terms of speed where there is a high number of substitutes for any given word. For instance, I want to use this to store common misspellings; expecting to have 4-10 variations of a word.
<expansion>
<sub>administration</sub>
<sub>administraton</sub>
<sub>aministraton</sub>
</expansion>
When you run a fulltext search, how does performance degrade with that number of variations? for instance, I assume it has to do a separate fulltext search performing an OR?
Also, having say 20/30K entries in the Thesaurus xml file - does this impact performance?
No, but performance testing is very often quite system-specific. I'd suggest putting together some sample data and running your own test cases is your best bet.