NTP Service not setting time - running manually works

Installing and Using OpenWrt
18

"Conversation" starts here: NTP issues on a dumb AP configuration - #13 by Cthulhu88

What I am using right now for my dumb AP and my dumb switch: have my main router host a NTP server (main router doesn't exhibit the problem), then have all dumb devices pull NTP data from the main router via its raw IPv4 address 192.168.0.1.

Also, something that I haven't mentioned in that post chain is that the problem is very likely caused by a bad /etc/resolv.conf, not necessarily failure to resolve names.
See my main router setup:

dnsmasq writes a local resolv.conf at its service start:

search intranet
nameserver 127.0.0.1
nameserver ::1

dnsmasq forwards global internet requests to dnsproxy:

server=127.0.0.1#5353
server=::1#5353

Because dnsproxy spams my syslog with large network unreachable errors until both wan and wan6 are up, I only start dnsproxy via the following hotplug:

INTERFACES="wan wan6"
STATE="/var/run/dnsproxy"

check_interface() {
    local iface

    for iface in $INTERFACES; do
        [ $INTERFACE = $iface ] && return
    done

    exit 0
}

all_up() {
    local iface

    for iface in $INTERFACES; do
        [ ! -f "$STATE/$iface" ] && return 1
    done

    return 0
}

case $ACTION in
    ifup)
        check_interface && mkdir -p "$STATE" && {
            flock -x 1000 && touch "$STATE/$INTERFACE" && all_up && {
                service dnsproxy start
            }
        } 1000>"$STATE/.lock"
        ;;
    ifdown)
        check_interface && mkdir -p "$STATE" && {
            flock -x 1000 && rm -f "$STATE/$INTERFACE" && {
                service dnsproxy stop
            }
        } 1000>"$STATE/.lock"
        ;;
esac

exit 0

This leaves DNS unresolvable for about ~10 seconds, but sysntpd doesn't deadlock and works correctly after DNS services become available.