We have run into a problem with build failures at high concurrency using -j12 on Solaris. The problem does not occur at lower levels of concurrency.
This problem has been traced to a coprocess ksh spawned by a recursive nmake. Apparently this ksh is exiting prematurely, resulting in a SIGPIPE signal to the parent nmake process. This occurs with the ksh88i supplied with nmake on the Solaris platform.
After discussing the issue with David Korn we recommend using concurrency levels of 11 or less on Solaris platforms. If you must go greater than 11 please contact technical support.