I would assume (given my limited scope of knowledge) that it is incredibly hard to gauge when a game will be "completed" (in quotes because very few games released today are truly complete, i.e. polished and ready to be released), and there are a lot of variables that go into selecting a release date that are not related to the creative side but the economic side; marketing is totally different, because--not only do you need a quality product--you need to look at your competition and what market trends are for the period you're looking to release in (e.g. Holiday 2012, etc.)

For example, if you were a relatively unknown developer who was looking to get its game out before Black Friday rush, when would you choose to ship it. Because of juggernauts already on the scene, Halo and Call of Duty (to name a couple), your game is likely to be looked over more so if you release around a similar time. However, because of the small window, when is the correct time to release? Before the release date of more popular titles? A couple weeks after? It can be really tricky.

Obscene hours and luck, I remember reading a book about game history and it had an interview with Eugene Jarvis (guy who made Joust) and according to him he got Defender up and running like 5 minutes before he had to present it. (I'm saying he was back stage fiddling with the thing right before it was unveiled)

As a HUGE Jarvis fan, I can add some info to that. First, they burned the ROMS to the hardware the first day and then it didn't work. Well, they plugged them in backwards and blew them. Then they took the game to the show with some working ROMS, and it was there. They worked on the demo mode while the machine was at the show running as they didn't want their game to be the only one without a demo/intro screen. It only took them a few hours to program that, after which they updated the ROMS at the show one last time.

Defender is one of the best works of art ever made. Such an amazing game too! I'd never be able to work with such a screwed up memory map. I even know the first opcode by heart. A clear $D000 command to clear the watch dog circuit on the PCB. :']

Speaking as an NES homebrewer with a completed title, the way I got it done "on time" was to dramatically overestimate how much time it would take me to learn to program for the NES, create tools for making a game, and then proceed to make a game. I estimated about 5 years initially, in late '08. It turned out to take about 3.5 years. So that's how I finished "on time" and indeed ahead of schedule. Also, I don't work anywhere close to 8 hours a day cause what I do isn't really a professional operation. We work a few hours precisely 3 times a week, without skipping a beat. The nice thing about this is we can continue to lead normal lives, have other interests, interact with family and friends, etc, but still complete games, and never, ever burn out even once (hence, tortoise and the hare symbol for Gradual Games). NES is a particularly nice platform for this because it will never become obsolete. Tools available for it only become refined---they never stop working or go completely out of date (except NESticle haha).

I'm in the upper half of a huge NES project. I rarely have free time for it now; I work full time and can't work on it at home much at all; I have three time windows in the day to work on it: on the bus to work, during lunch, and the bus back home. It adds up to 2 1/2 hours a day, giving me roughly 12 hours a week. Mostly programming and testing is done during this time. Artistic elements are done when other free time suddenly arrives, a day at home to myself, whatever. Whenever I happen to get the opportunity,I take it.

They set an appropriate timescale for the work they need to do, then do it. You only need to be working like a slave if you're short on time or have deadlines to meet.