Ok, there are 20 different types of bias. I'll list the ones that a poll and a handful of threads being used as a good sample are subject to.
1 anchor bias
An over reliance on the first piece on info you hear
The poll is subject to that as the first line in most of the polls is "there is a cracking issue". Very few start "there is no issue"
2 availability bias
Overestimating the info available to you.
So a few hundred maybe people write about a cracking issue so most of the available data says that. But that's because you don't have data from the 100'000s of people who don't have any problems.
3 bandwagon effect
When one person is vocal about a view in a convincing way he creates followers, who are also vocal. This is how extremists, suicide pacts etc happen.
So everyone with a crack is understandable annoyed. Then someone shouts "hey if we get a petition you'll all get new P3's with more efficient motors and new shells" great! Of course poeople pile in! I'd love one to (I also have the first gen shell and old motors) do I think I have a right to one based in the fact I have no problems and it performs as advertised and to the specs listed? Of course not!
4 confirmation bias
Giving more weight to the results that confirm what you want to believe and ignoring those which discredit it.
Eg calling anyone who states they don't have problems a DJI fan rather than giving their result equal importance to those that don't.
5 selective perception
Allowing our expectations to perceive how we view the world.
Let's be honest everyone starting one of these polls expected it to prove there was a problem. I doubt any of them set it up because they genuinely weren't sure and wanted to find out.
Then there's the sample (people that took part) these should be picked as randomly as possible. The problems with this were:
1 they weren't picked! They selected to take part. Someone with cracks WILL take the trouble to "vote" as they want to moan, air there annoyance and hopefully get something out of it! People without crack won't, as they wont get anything out of it so they won't give a toss.
2 this hope that all pre a certain date models will be replaced would mean even people without cracks may say they do have cracks as they may get a new P3 with more efficient motors out of it.
3 How many people with a crack do you think found the poll googling searching "P3 crack"? Now how many people without a crack do you think googled "P3 crack"
4 how many forum users with a crack do you think opened a thread called "P3 crack"? How many users without a crack do you think opened the same thread?
Finally your sample size is way too small to be statistically valid. I don't know how many P3's have been sold or how many people answered the poll, if you do plug it in here.
Sample Size Calculator - Confidence Level, Confidence Interval, Sample Size, Population Size, Relevant Population - Creative Research Systems
I use 95% for most of my projects. Now let's assume just 10'000 P3's have been sold. To be 95% certain (+ or - 2%) your getting a true representation from your sample you'd need to get answers from close to 2'000 people. That's assuming of course you correct all the above biases and sample selection problems.
So Ayrtonginster, that's why I think the poll is absolute rubbish and can go in the bin.i hope this explains it.
To your other points the P3 is the only DJI product I have ever purchased so I wouldn't say I'm blinded by loyalty.
You state 1 in 3 P3's have cracks, I'd love to see you post where you got that data from!