RFE: rethink initial jitter.
RFE, for after 1.0
Setting the hitter to zero on startup has perverse effects.
For example, if a refclock was not available at startup, it jitter stays at zero. Other working refclocks and servers eventually get valid jitter values.
If, after startup, the failed refclock comes alive, it has a jitter of zero and will outvote other, more stable time sources.
I see some code that tried to set the initial jitter to syslogd(precision), but later code seems to zero that out.
A similar issue arrises when the system clock is changed on startup. That time jump is added to the jitter of all the refclocks and servers, throwing off the jitter.