If I wanted to try this out, I probably wouldn't be starting from there
I don't see the HH5B as being a useful tool in "seeing what happens" to the line - so I would almost certainly swap to a model that gave me access to all the stats. However, I don't use BT Infinity, so I'm not sure what limitations that ISP may inadvertently place on your connection.
I've used unlocked HG612's modems with a broadcom chipset (in combination with a standalone router), and a Billion 8800NL combined router/modem with a newer broadcom chipset.
Once you've got to the point that you can see all the line statistics, then you're in a much better position to see what happens.
As it stands, I suspect that G.INP is active (though you can statically figure out what cab type you have to see if this is likely), and is much less likely to then add draconian DLM interventions through a few power cycles. You might need a *serious* number nowadays ... but I've not heard anyone try deliberately since G.INP went live.
Still want to do it? I'd go for 5-6 line disconnects (not power cycles), and see. Then a burst of a few more if needed.
As it stands, you might not see a drop in sync speed. Your attainable is already 10% higher than the package sync speed, and that might be enough to mask the FEC overhead when/if DLM turns it on. You're probably more like to see a change in latency first.
Talking of latency, I'd use a "ping monitor" or "BQM" from thinkbroadband to monitor latency. However, this is only worthwhile if you have a static IP, and your router responds to pings. I'm not sure the HH5 does that.