Ok, I know the following is absolutely my fault, and I'm really glad nothing happened to the P3, but I'd be really glad for some input on why that happened, and how to avoid it (beyond never doing it again ).

For the past week it has been raining incesantly, and next weekend it's supposed to rain some more. Last night there was a period of two or three hours, where there was no rain. Thinking that the batteries might suffer from sitting around charged for such a long time, I decided to go down to the park, and just let the P3 hover for 20 minutes and get the batteries down to 50%.
When there, the imp of the perverse (to quote Poe) made me think: Why not test your low altituted piloting skills while you're at it?
Great idea... not. Started the machine, flew it to ~75 cm high, and I noticed that, for the first 10 or 15 seconds, it drifted sideways instead of hovering. No compass error, and after those 15 seconds, it stabilized nicely. I pressed forward, and I kind of felt that the P3 was slightly strafing towards the left, instead of going fully forward. Deciding to ignore all warning signs, because, you know, I was being a bit thick, I got to the entrance of a small roller rink, around 1.5 meters wide, with ~1 meter tall metal poles on the sides. Of course, the P3 strafes towards one of the poles, hits it with the two forward propelers, and falls on its back to the ground. I panic, and instead of immediately doing a CSC stop, I try to stop it by pressing down for three seconds (doesn't work).
After I collect myself enough to do the CSC, I pick up the machine,  and check everything. The giant spaghetti monster be praised, the gimbal is all right, no dents on the arms; only the propellers took a hit by spinning on the ground! Tested it later, after drying it, and everything is running perfectly.
So yes, it all ended up as an incredibly scary, and stupid, mistake on my part, luckily without any serious consequences.

Now, my questions and/or theories on this drifting behaviour.
Do you think this is more of an improperly calibrated compass, or suboptimal VPS conditions?
Can it be because it was nightime, or because the floor was wet (i.e. darker)? Will the VPS have trouble under those conditions? The floor was tiled, but these were textured, square tiles, will the regular pattern override the texture?
Either way, I'd still be really interested in flying at low altitude under conditions of moderate lighting (i.e. forest paths during daytime, and things like that). If I decided to do that, what kind of recommendations can you give me(beyond "Don't be an idiot, you allready know what will happen! Do not, under any circumstances, do that!")? What would be the safest way to train?

Thanks a bunch!

The VPS doesn't work anywhere, anytime.
It has requirements .. Operating Environment - Surface with clear pattern and adequate lighting (Lux > 15)
You said it was night.  
Was the surface well lit?
Was the pattern clear enough?
Start out somewhere that does meet the necessary conditions rather than somewhere you're not sure of.

I turn VPS off for all my outdoor flight. Also, given that this is a drone with a camera designed for distance shots (not skimming the ground at 2 feet high), I don't feel the need to train at very low altitudes. It works fine without the VPS on - maybe I fly 1 meter off the ground at the lowest.

Thanks for the advise! The degree of lighting has always been a concern of mine with the VPS... The main problem being that, because of my job, and living so near to the equator (i.e. 12h of light, mostly working, and 12 h of night), 90% of my flight time is at night. I'll look into getting a lux meter, and will definitely use it if I'm going to be relying on the VPS. Either way, I'm pretty sure I'll be avoiding anything like this again.

Also, I've been reading in other threads that people are experiencing side-ways drifting possibly due to deviations on the IMU calibration, even though the app itself might not be giving such an error.
http://forum.dji.com/thread-20767-1-1.html

My question on this, is it possible to do an IMU calibration indoors? Or should I follow similar rules to the compass calibration?

In reading your description, I get the following feelings:

Thank heavens you didn't destroy your bird.
Flying in a more open area for the first time would have been a good idea.
(It's obvious you know that now!)

The tiles and other surfaces you mention - it had been raining (a lot) and virtually all those surfaces were wet - did that automatically make them "...a highly reflective surface?" Strong cautions have been issued about the Vision Positioning System (VPS) difficulties - one that stands out is the "...highly reflective..." issue. I don't know the technical whatever about the VPS, but flying over calm (reflective) water, blah, blah, blah.

There's a little bit of info in the manual about this - but not much. I think most of what I've seen/heard thus far has been through several forums. That was sufficient to instill VPS-fear in me - I turned mine off. Don't fly indoors at all. And, like droneflyers wrote - I tend not to go skimming along the ground either.

I've seen a few shots here and there where someone shot some moderately long videos about four or five feet of the ground. Didn't buy my Phantom to take pictures at the same height I could do with a hand-held camera!

aburkefl@gmail. Posted at 2015-6-15 23:28
In reading your description, I get the following feelings:

Thank heavens you didn't destroy your bi ...

The more I think about the whole thing, the luckier I feel. As you say, off goes the vps . From now on, no more skimming... at least 3 meters off the ground. Thanks, though, for all your input.
p.s. it was not my first flight, though... just my first one without the support of an actively thinking brain, apparently.

Where is DJI ?    They should honestly provide proper guidance on using VPS and flying indoors.   They want us to do all their beta testing for them?   Shame on you DJI.

kenmasters Posted at 2015-6-16 00:15
Where is DJI ?    They should honestly provide proper guidance on using VPS and flying indoors.   Th ...

That's a good question... even though in my case, there was more than enough fair warning that something was wrong (I'm really not blaming DJI here), a bit more of a realistic explanation of how/when the VPS works would be nice...

Here. Rince, sorry for your accident. My two cents, it is less of the suboptimal VPS conditions. If the working conditions are not met, the aircraft will drift, but from your description I suppose it was almost out of control when it hit the pole.
Ultrasonic can be absorbed by water, and low light condition can also affect the performance of VPS system. So pls do some further tests if you wanna fly on the forest path.
If you wanna know more about what the accident is leaded by I think you can extract the MC data and send it to support.

DJI-Autumn Posted at 2015-6-16 16:47
Here. Rince, sorry for your accident. My two cents, it is less of the suboptimal VPS conditions. If  ...

Hi Autumm!

Thanks for the explanation. You are right, up to a point. When it hit the pole, it was clearly drifting by itself, but if I'd been a bit quicker I could have pulled it back. Either way, yesterday I did the update to the new firmware, an IMU calibration, and today a compass calibration. The result is, again, very nice and steady positioning, without any drifting.
One question, though. I just watched my first videos after the update, and noticed that the flight data output has changed. Instead of speed and so on, now I get barometer and ultrasonic. I suppose those mean altitude as measured by VPS, and by barometer. Looking at those, I get wildly different values. Can you comment on that?
I'm attaching a screenshot for clarification.

Thanks very much!

Good to hear that DJI is listening.   I can only hope for some formal communications to guide proper/safe flying with VPS.  

I had an similar but unrelated question.   

How do I keep VPS on all the time as the preferred ground tracking method?     There is an option to turn VPS off, but is there an option to shut off GPS ?   As I understand, even with ATTI manually on the controller, if the phantom sees satellites in goes in P-GPS mode.   This throws off VPS or does it ?   I tried flying indoors and it was drifting.   I'm near windows and there is GPS.  Don't know if that would cause any stability issues.  

Rincewind Posted at 2015-6-16 17:19
Hi Autumm!

Thanks for the explanation. You are right, up to a point. When it hit the pole, it was ...

Hi Rince,

Did you upload the screenshot?

DJI-Autumn Posted at 2015-6-17 11:47
Hi Rince,

Did you upload the screenshot?

Yes, sorry... here it is.

DJI-Autumn Posted at 2015-6-17 11:47
Hi Rince,

Did you upload the screenshot?

DJI-Autumn Posted at 2015-6-17 11:47
Hi Rince,

Did you upload the screenshot?

Yes, sorry... here it is.

DJI-Autumn Posted at 2015-6-17 11:47
Hi Rince,

Did you upload the screenshot?

I'm sorry, it looks like firefox and the dji forum are having trouble.... somehow the attachment doesn't appear in the text. Basically, the barometer is saying 3.6 meters, while the ultrasonic says 1.5.