We were ingesting one level2 Nexrd2 site from upstream, and it was
adding up to 2 mins delay, with data arriving in "bursts" consisting of
multiple volume scans. That was a simple one liner in the ldmd.conf.
It seems that the more data flowing, the less latency and "bursting"
occurred. In any event, this upgrade fixed the issue and its working
fine now. Thanks Steve.
Yes, 6.6.3 is the "Goldilocks" version: pre-6.6.0 flushed too much,
6.6.0 through 6.6.2 flushed too little, and 6.6.3 seems to flush just