In this case, the formatting of the bits into packets don;t matter. Speed experiments are done using a PRBS pattern to assure an almost random set of bits. The point is to prove that data rates are possible. What those bits represent is irrelevant here.
Is your question about fiber properties or VCSEL poperties?
In general, 1310nm VCSELs have lower ouput power and, to date, much lower modulation speeds than 850nm devices. In the literature, 30Gb/s over 10km of single mode fiber has been demonstrated from a 1270nm (~1310nm) VCSEL.
For 1310nm VCSELs, the single mode fiber distance will be in the 10km range, limited by either the available optical power or the spectral width.
While reading this article, a question just occurred to me, which I came across a few days back and I did not know the answer. To implement Fast Ethernet or Giga-bit Ethernet, we generally use the off-the-shelf processors with in-built MAC, 0ff-the-shelf PHY chips compliant to the Ethernet standard and magnetics, connectors per standard. What are the test cases which need to be done to check whether Ethernet implemented actually complied to the respective standards?