Time constraints were the biggest one. 18 months to make a game the size of F:NV was stressful and difficult. I don't really think console limitations were a big deal, but our lack of familiarity with the engine made it difficult for us to optimize. https://twitter.com/SmokeyTheTear/status/1340349710205399040

Thankfully, Bethesda internal dev did help us with some optimization later on. The toolset/pipelines for Bethesda's engine are fast, the fastest I've used, honestly. There is no other engine I can think of that would have allowed us to make that much content that quickly.

Two of the biggest hurdles were technical landmines that appeared relatively late in development. First, Havok Physics needed to be upgraded several versions because the version used in F3 would not compile with date-compliant Sony SDKs. Our programming team was very small.

We needed to bring in programming resources from other teams and from outside the company to upgrade and bugfix Havok sufficiently. I say "sufficiently" because obviously there were still physics bugs at launch.

Second, we needed to fix a large list of outstanding TCR/TRC bugs that we were unaware of until pre-certification. Havok and TCR/TRC compliance took up the lion's share of programmer time in the last few months of development, meaning gameplay bugs went unfixed.

And sure, the PS3's split memory pool was a pain in the ass, but we knew that going in. Charlie Staples (Lead Area Designer) did a great job coordinating memory management and the area design team as a whole did what was necessary to just make the game run.

That's why the Strip, Freeside, etc. wound up being split up late in development. The areas just wouldn't run as we had originally built them. The team learned a lot of good lessons from it that they were able to apply to the DLCs.