I guess it's just that the information you've provided so far is way too unspecific or vague to allow for any serious recommendations.
So you are getting either partial answers for certain circumstances of your observations that are clearly definable (as e.g. @jfb's explanation of how TTLs cause websites to show ads even when Pi-hole is re-enabled) or just wide guesses at less related simliarities to some future features.
Let me have a try at another, a bit more elaborated guess by taking your most recent post into account, and by taking the discussion back to your starting point: Your topic's heading - Blocked sites without any info in the log.
Now, Pi-hole wouldn't block anything without letting you know.
As you don't see anything blocked in the logs, Pi-hole doesn't block it.
Indeed, you further correlate DNS queries that Pi-hole is showing as successful with pages that do not load.
In that most recent post from just a few minutes ago, you provide evidence of two additional noticeable facts:
- a DNS query (related to a page not loading) takes an unusually long time to finish - well over two minutes.
- you are using DNSSEC
If you happen to run Pi-hole on an RPi, using DNSSEC might be the reason for 1) to take so long:
RPis are lacking a real-time clock (aka RTC) used to keep track of a precise time, so they have to re-sync with internet time servers way more frequently. As DNSSEC relies on exact timings, a resync hitting your RPi will also delay resolution of hostnames for that time, or may even prevent sucessful resolution completely.
This failed or delayed resolution may in turn cause your browser to give up on waiting for a hostname to resolve or just waiting endlessly, resulting either in a time-out or some (significant) fraction of your page growing stale.
This explanation would fit in nicely with your inability to put a certain time or a certain domain to your problem that would guarantee reproducabilty of your observations.
On the other hand, it doesn't fit quite well with you still observing strange behaviour even when Pi-hole is switched off completely for a longer period.
The best approach to further isolate your problem would possibly be to
a) forego using DNSSEC for a while and see whether you still experience problems
b) manually configure your problem client (iPhone) to bypass Pi-hole to see whether its divergent behaviour is at all related to Pi-hole
I'd leave it up to you how to best combine or separate these two options 
In case you'd need further assistance in dealing with your findings, it'd be probably helpful if you could post your debug token here and provide any interesting additional information about your network configuration.