I would imagine most people are taking discrete samples every now and then with a PH pen that can measure in 0.1 increments?
I measure every 3 seconds using a sensor than can measure in 0.001 increments (stable accuracy +- 0.005, 3 readings averaged to reduce noise giving a more accurate reading every 10 seconds). As such a world of new information has come to light regarding a lot of things that happen in the tank.
When I use a doser and graph my PH when I manually add some acid to a 100l tank of tap water, (PH 8.1, fairly hard) at the point the acid is added the PH drops to the lowest point it will in less than a minute or two as the acid mixes into the nute solution. It then bounces back, over up to 3 hours, to a maximum (sometimes very) close to the original value. It then slowly settles down to the actual value the acid has adjusted it to, this whole process can take up to 12 hours. This final value is usually quite close to (but normally slightly higher than) the value reached in the initial PH drop, just after dosing.
I find that the exact time between dosing and taking a reading can put your reading out from the actual 'settle value' by almost the amount you have dosed. I find that the larger the dose you use, the larger your margin of error can be. As a result of seeing this, I now micro dose my nutrient tank. The vastly smaller dosing (around 1-2ml at 1-20 dilution) reduces the variance in PH and often the slower rate of change is one that can be overcome by the effect the plants, irrigation etc have on the PH. I find this reduces dosing errors to so low they can be ignored.
Even with just a 100l tank, when the solution has to interact with roots and pipes etc it can take 7-12 hours to settle down from a dose that would take the PH from 7 down to 6.
So, as an example, with my system you could dose an amount of acid that would take your tank from PH7 to PH6 at midday, measure 3 minutes later when the solution has mixed and you would read around 6 (usually just slightly lower than 6 but you would not see that with a pen). If you continued to take readings over the day you would see the PH rise until around 3pm, when you would see the PH be around 6.7. If you continued to take readings you would then see the PH drop down to around 6.2 by 6pm, and it would then settle at 6 by midnight. As you can see, if you take these readings as true and adjusted your PH at 3pm down to 6 again, your final PH would be closer to 5.3 and it would reach this value while you were asleep. As you can see, this would lead to 'weird' PH fluctuations (in my system at least) caused by not taking readings at the right time and dosing at the wrong point. After seeing and graphing the interactions between acid and tap water to a high accuracy my advice is to measure PH 3 minutes after dosing (or at the lowest point it gets to immediately just after dosing), or 12 hours after dosing, when everything has had time to settle, not in-between. Then you will accurately measure the actual PH your tank will end up at.
I find when my nutrient strength is higher this effect is lower as the nutrients often contain a PH buffer which tries to keep the PH closer to 6 than over 7. This effect is also very much reduced when dosing pure RO water with no nutes. That reacts almost instantly (<5 mins) and does not take hours to settle down.
I also find that with tap water lowering the PH from >PH7 to <PH7 takes a lot of acid compared to changes under or over PH7. I suspect this is due to the buffering going on.
This is with my system, the local water, and the acid I get from my local hydro shop.
Your mileage may vary.
I always dilute my PH down at least 1:10 usually 1:20 so this type of problem is less of a problem.
If, like my system, it takes a relatively large amount of acid to move from 7.1 to 6.9 but far far less to move lower then you would have the effect you experience. I add that sort of amount to my 100l tank to shift the PH down 0.1-0.2. A 5l bucket would show 20 times the change, being a drop of 2-3 PH points so a drop from 6 to around 3-4 is the change I would expect.
A few ml to drop 0.5PH in a 5l bucket would be at a dilution of 1:10.
I would say give that a go or to give you more of a margin 1:20. You can always add a bit more, you can't add a bit less...