I'd say a few things about this. First, we need to realize that work requirements and program generosity are two separate (although related) issues. Certainly work requirements are going to limit access to benefits (that's the point), but funding levels can be such that these programs don't have to be rationed in addition to the rationing implicit in the work requirements.
The other concern with work requirements is that if labor demand isn't there the changes from AFDC to TANF aren't going to do low income families much good. This, I think, is where the TANF subsidized employment program can really help and why it's a real tragedy that came to the end so quickly. Other strategies like hiring subsidies may work too. My research last fall (which I'm looking to submit somewhere this spring) was less positive on hiring subsidies, so I'm not sure how good of an idea this is anymore, but (1.) although I think my research was solid and an improvement on a lot of existing studies, it's just the beginning of my work on this, and (2.) there is a lot of more positive research out there too.
The point is, the ability of TANF to act as a safety net is constrained by at least three forces:
1. Funding levels
2. Benefit generosity, and
3. Access restricted by work requirements
We can fix 1 easily if we want to, and we have strategies for fixing 3. Once that is done I'm not sure TANF is a failure, although we may have to revisit 2. There were good reasons for pursuing welfare reform. The product wasn't perfect but that doesn't mean AFDC was either.