It seems you're making this very hard by having selected European Spanish. A language like English has a lot more homophones due to the messy spelling, while Spanish has much more consistent spelling (generally a good thing) making it harder to design a study like this, and more importantly possibly meaning that speakers would have fewer cognitive effects of this type, while English speakers are known to have major cognitive effects based on organizing vocabulary based on spelling. (Actually testing Spanish like this could be a good comparison for that reason but doesn't make the experiment easier to design.) Furthermore, using Latin American Spanish or other dialects (including some in Spain, such as Andalucian), would allow you to find many more orthographic mergers with c/z=s. This is a known topic at least anecdotally for testing spelling of some, especially colloquial, words in Latin America, such as the difference between voz and vos (where that informal pronoun is used).
That said, I will let you know if I think of any homophones. Have you thought about any English borrowings or brand names? (That might introduce additional complexity for the experiment, though.)
Finally, why not take advantage of something Spanish has not found in English, specifically the use of accent marks? That is, you could compare hablo/habló, or similar pairs (maybe where the meaning is more different, not just grammatical) to study something that can only be tested in a language where that is a factor. And many Spanish speakers don't bother writing the accent marks on the internet, for example, so it might be interesting to see how they process the differences. (I want to emphasize that I think experiments not studying English are great, because there's such a research bias for English; I'm just thinking about whether something more applicable or common in Spanish could be included.)