Route based on latency?

All,

I need to create a self-contained package for a client of ours who will tour with a recurring event. Their guest management system relies on a timing-critical database application running somewhere in a cloud. Due to recent bad experiences with venue internet and partially blocked ports, the package will also include a LTE client.
From there I will set up two SSTP tunnels (one through venue provided link, one through LTE, SSTP becaus of venue firewall compatibility) back to our own core and access their cloud services from there.
I currently can’t see a way to judge link quality based on latency (so the tunnel with lower latency is always preferred) Has anyone done something like that before and could give me a smart hint?
Due to budget reasons I want to avoid deploying a SilverPeak vxoa appliance on that system…

Thanks,
-Chris

Just a though: why not use “normal” dynamic routing protocols, but where route weight is (partially) based on latency?
Something like this: https://docs.aws.amazon.com/Route53/latest/DeveloperGuide/TutorialTransitionToLBR.html

another relevant hit: https://networkengineering.stackexchange.com/questions/2765/using-latency-to-calculate-ospf-metrics

Well with the Mikrotik you do have some fancy scripting options.
There are quite some examples where the box has 2 links and with a script the default-route is switch is 1 of the links is down.
Offcourse, here we want to play with the response-time above X ms , then switch to other link. Partial example is below in the link.

Something like this, but then enhanced with some code to alter the default route depending on the amount of “ms” response of the gateway or something.

http://forum.mikrotik.com/t/avg-rtt-from-ping-command/75110/1

So I guess you could try to combine some scripts to switch tunnels/routes based on RTT-values etc.

Thanks for the hints.
I already considered the scripting way but was unsure if there was something more obvious that I might have missed.
-Chris