You still need to test all devices to make sure, its malfunction which causes draw. Unless you test the device you don't know if a eBay purchase gave you a partially fried soft switch pass transistor that drains 5 to 50 times the expected quiescent current while powered off!
Any battery will leak when over drained. It leaks more when you drain it fast but the idea is.. you should replace them at a reasonable voltage under nominal. If you are still expecting good chances of not leaking because your device decides to extract energy from them at 0.8 volts.. its not a good idea to try to squeeze that out of them (if the batterizer actually worked well, it would cause mayhem in terms of battery leakage). What the premium batteries are supposed to do is offer more useful life for power devices in the nominal region (extend that platue of 1.2V). They do not vent the gas and as the electrolyte is spent the gas pressure rises...
As dave jones has shown, using the batteries under nominal.. does not save much. There is not much capacity left there. It is not worth to squeeze them dry for the extra 5-10% (if that, usually at this voltage range the device starts to 'suck' by having a dim display, being quiet, etc.. so its essentially crippled. You could say for the most part you are squeezing almost nothing out of the battery, and you are also putting up with a improperly functioning device in terms of UI, accuracy, etc. Even good DMM and stuff often have low battery voltage related errors).
And try lighting a wall clock on fire with a small short, it will just smolder. but a lithium flare will ignite most any difficult to ignite fire retardant material. I did that test on a plastic table and it melted through the whole thing and set the corners of the burn hole on fire with dripping plastic. Those normal AA devices are safe because they don't have what it takes to actually get enough plastic hot enough to start a fire.. but I am not kidding, the lithium primary battery in AA size is similar to a nautical flare once you get it started. The normal road flare you see is pretty tame because it does not have flammable metals in it, just sulfur, sawdust and red oxidizer. Once you put flammable metals in there the flame temperature increases drastically, it starts spewing little fireballs like a sparkler fountain and flame/burn chemistry changes because of the refractory temperatures achieved. And that fire just gets more angry when you pour water on it and starts spewing more white hot fireballs. (its a end of the world volcano effect, the worst electronic component failure I know of).
A123 and such batteries also do it.. they double as decent fireworks and produce about as much smoke as a similar sized firework.. smash them on a rainy 4th of july for fun. The burn rate is more impressive then even a large sparkler.
Do actually try it, the videos are not very good at conveying the true story of reactive metal fires.
So after 10 years turned off, the device should not get under 1V ideally. If you get down to 1V after 6 months because of high quiescent, the battery is dead, has almost no energy left (the integral under the curve on the voltage range of interest is energy), and it is pressurized. Leave it pressurized and dead for 2 years and of course it will leak. The time constant for checking the devices is too frustrating, it should be at least good for 2 years by design IMO. At 0.8V you have like <5% capacity. No point in leaving it in. When you find that you should be anxious about how little time it has left and how its not going to be available when you need it.
The date mark of those devices is 'best use by' not 'in service until.. the service use date depends on the voltage/capacity/gas pressure. Of course if it leaks in the package its a QC problem. Not saying there is not bad brands. No one wants to design it because someone is stupid and leaves a battery that will not even power up the device in the circuit for 3 years.
I recommend checking the batteries yearly with a DMM and tossing most stuff that is under 1V because its not worth the money.. the value is still good if you don't squeeze 5% more out of it. Chances are you will just be late to work when you forget about a clock anyway, not worth the risk. Optimally test in circuit, but its difficult in some devices like flashlights, so then you need a battery load tester. Thankfully its like 5$. You don't need to be super precise because everything is cheap.
I would go with the gut instinct and cut down that 'good till' date by 30%. Have you ever seen any consumer device ever that has totally accurate specifications? Any moderately smart individual will do things like 'try not to use that drill on max all the time' unless they paid 500$ for a boutique power tool that retails for 250-300 from common manufacturers (in terms of power tools). I.e. plowing tons of holes in reinforced concrete with a standard hammer drill with the max drill bit size all day on a round robin shift of batteries..