There's a landmine of bugs in even the best EDA software. I'd just like for the software to run nicely once, like TensorFlow... sigh.
If you're thinking getting into this field, I might suggest a modular software architecture where the non-proprietary parts are open source and everything NDA-sensitive is kept closed source (pretty much anything the foundry touches/provides).
There are other problems too. Just in general, lots of bugs. For example, we'll run something and we'll get an obscure error, and then we'll ask the EDA company, and then two weeks later we'll get a generic "we'll look at it".
The hilarious thing is, sometimes the more expensive EDA companies aren't necessarily the better ones, our current partner isn't necessarily known as the most expensive tool suite, but we've actually had (so far at least >.<) had a much better experience with them than the bigger vendors.
Is your company big? Usually bigger EDA vendors don't care much for smaller design houses.
Btw. how are your experiences using Xyce? Do you get much speed up using its parallel processing capabilities?
Xyce is new and has some syntax differences, so it's not a mature software quite yet, so there are some troubles there. But as we've been starting to make it work, it's given us some amazing speedups on large clusters.