Method sysGetMilliTicks in solaris/hpi/src/system_md.c has possible problem:

long
sysGetMilliTicks()
{
struct timeval tv;

(void) gettimeofday(&tv, (void *) 0);
return((tv.tv_sec * 1000) + (tv.tv_usec / 1000));
}

The multiply of tv_sec by 1000, this will wraparound every 24 days.
(24 = (2**31)/(24*60*60*1000)) Places that measure an interval by calling
sysGetMilliTicks() twice may get a very large positive value the first
time and a very large negative value the second time, leading to all
sorts of interesting results. Note that the wraparound occurs every 24
days starting at some arbitrary time, not after the JVM has been up 24
days.

NOTE:
It does not occur after a JVM has been up for 24 days. It occurs every
24 days, based on the number of seconds since the EPOCH (Jan 1, 1970). So
the wraparound can occur between any two calls! Of course once it occurs,
it will not reoccur for another 24 days.