The colours vary per PPU (read: per NES console). Really.
It's been discussed before. (see Quietust's response). Grey is the most common, but I think every single person here has seen variance (some shade of green, peach, or red).
As for the duration: on the NES side, it's consistent (I believe every 1 second), but the real-world results (on screen) are going to vary because of what rainwarrior said. CRTs tend to handle video signal being dropped (line held low) quickly, so visually you see the screen flashing every 1 second. LCDs and other present-day TVs don't tend to do this, they tend to "screw around" a lot (often for several seconds) when the video signal is lost and restored; it usually has to do with logic relating to determining signal "type" (ex. 240p vs. 480p vs. 480i). So you might not see anything on Model X of a Samsung LCD, while Model Y of a Samsung LCD might show something, and Model B of a Sony might behave similarly to a CRT. There is tremendous variance that you have no control over, sadly.
If you're developing something that requires you to have specific timing or are concerned about this for a good reason, I would suggest using an actual CRT for testing/development, rather than fighting with the idiocies of LCD behaviour. Another possibility I suppose would be to use an oscilloscope instead of a TV; I imagine lidnariq could talk about that (I sure can't).
Maybe it would help us if you explained why you're asking the question. It's a good first time post and legitimate question, but your follow-up questions are perplexing -- one of those "why does this person care about that" scenarios. If we understand better the "why", maybe we can help you better. :-)