[exim] System time correction during or after Exim startup a…

Hi everyone,

Not sure where else to report what I've been troubleshooting, but I thought
it worth starting here.

For years now, I've been running Exim as a mail relay on CentOS 7 AWS
instances. Those instances have been whitelisted by Amazon for this purpose
and under normal circumstances operate perfectly. I'm currently running
Exim 4.94, provided by the EPEL package.

In recent months, I've received complaints from users about slow delivery
times for email transiting the servers after instance restarts. Log
messages show higher queue times, and they all tend to be consistent with
one another. For example, the QT=3m17s entry might be logged for all
processed messages following the instance reboot. If I stop/start exim, the
queue time returns to normal - approximately 1-2 seconds per message.

Today, I discovered that NTP is stepping the time pretty significantly when
the service starts on boot. In the case of the queue time cited above, NTP
advanced the system time by 184 seconds just after the NTP service started.
That fairly perfectly matches the 3m17s logged by Exim.

Have any of you encountered this issue before? I can certainly adjust the
Exim startup to wait for NTP to complete its work - that will eliminate my
problem - but I was curious to know if it sounds like Exim is functioning
as intended here.

Cheers,
Matt