Just in case it helps track it down, I had a way to reproduce this.
On sbnc1.2 I had a script that foreached through users and would tcl jump 4 users at the same time. Since I wasn't using oidentd there was no valid reason to have connects as slow as they were. Every time I used the command however, I would get a timewarp of somewhere between 5-10s. My theory at the time was that it was to due with connection blocking.
Side note: If I ever tried to jump more than 5 users there seemed to be a hit/miss chance of the bnc crashing completely.
Just in case it helps track it down, I had a way to reproduce this.
On sbnc1.2 I had a script that foreached through users and would tcl jump 4 users at the same time. Since I wasn't using oidentd there was no valid reason to have connects as slow as they were. Every time I used the command however, I would get a timewarp of somewhere between 5-10s. My theory at the time was that it was to due with connection blocking.
Side note: If I ever tried to jump more than 5 users there seemed to be a hit/miss chance of the bnc crashing completely.