You seem to be saying that you have poorly written code that runs quickly on Solaris 10 systems consuming huge amounts of CPU cycles while not checking for error returns on sockets and that the same code when ported to a Red Hat version 6 Linux distribution runs slow and stops when the same networking error occurs.

Without seeing the Solaris 10 code you're using and the RHEL 6 port of that code it is very hard to guess at what might be causing the differences in behavior you have described. My crystal ball isn't working well enough tonight to even hazard a guess.