Runtime of gpmetis with contiguous flag

Hello,

I wanted to ask if there is a (theoretical) estimate how forcing contiguous clusters influences the runtime. I did not find much information about that in the papers on METIS.
When I partitioned a graph with 8M vertices and 100M edges, no weights whatsoever, METIS needed a few minutes, but with the contiguous option, I had to abort it after several days. (There was enough RAM available.)
Moreover, are there plans to address this in the near future?

Cheers,
Markus