I think there is no hard limit.
You should avoid adding too much ED if you don't want to experience speed issues in your survey experience.

I've heard that as much as one hundred ED will not be an issue.

Do you need more than that ?
If yes, I think you should contact support team.

Hi, thanks for the reply! I was going to need a lot more -- 320 stimuli with 4 variables each. FWIW, I had done as many as 75 stimuli x4 variables so 300 embedded variables is not an issue. I did not see any speed issues with those studies but I was using the API, if that makes a difference.
But even making the embedded variable sets was going to be an unnecessary effort for 320 stimuli. I got around it by concatenating the output strings and unpacking in post hoc analysis. The max variable length is ~20K bytes and that was easy to fit under.
thanks!