Decision for timing closure based on seeds

The fitter starts by making a pseudo-random initial placement, which is affected by the seed. Thus, different seeds can lead to different final configurations with different properties (timing, resource usage).

When your design is on the edge, you may end up with cases where your design works and others where it doesn't work (doesn't meet timing or doesn't fit).

The cases which passed timing analysis will work correctly and you can use them.

So, you can say that you have achieved timing closure.

The problem is that since you're on the edge, any small changes you make may push you into the "fail" range.

Best case scenario, you'll have to run another sweep to find a seed that works.

Worse case scenario, you'll no longer find any cases that meet timing requirements.

It's not a comfortable situation to be in.