Geofram, you yourself say that online diplomacy is a different game to Face to Face. It's a different game to PBEM and PBM too, and must be viewed differently to both of them. When deciding on a feature, no precedent is appropriate, because Internet diplomacy of this kind is in it's infancy. As far as I can tell, phpDiplomacy has been pretty close to the first large diplomacy website, if not the first. Let's not think of this in terms of emulating any other version of the game.
You talk about the effect the finalise feature has on the game, and yes it does have one. Is this necessarily a bad thing? I don't think so, after all, internet diplomacy can be a bit lacking when you only have press to deal with- how can you tell the difference between well scripted lying and the truth on the internet. There are features of FtF that tell you more than what people say- their expression, who they speak two etc. This works in the other direction, but why in this direction is it necessarily negative? Players have to take care not to be obvious in speaking to the next ally before a stab. They need to take care not to be obvious in their finalising habits.
Now, you claim we have no choice- that is wrong, you can state at the start of the game "I am going to hit update on the first log-on after an adjudication to turn on the grey tick- it means nothing". You can work around it quite easily. You might be better off trying to use it to your advantage. You can even update "all hold" in this way and remove it.
How about the effect of being unable to enter orders? Well, if you can't log on, we would have known anyway, because we could see that you hadn't logged on. The problem is that you have to notice it just before finalising. What if you can log on, but don't enter orders? Well, again, you only have to hit update once, and you are done.
As for telling people that you have a tentative plan in place- surely everyone has a tentative plan in place all the time? Of course, you can, again, get round this as I described, should you so wish.
Finally, we have to know when people are on line to know whether they can have read and responded to our messages, there is no argument to be made there.
When Dunecat says, "One could argue that this is the most significant issue facing 0.9, so I'm glad it has its own thread." I wonder what his argument could possibly be, given how easily neutered this feature is, I find it remarkable that Centurian declare so unequivocally that "This is undeniably a HUGE change." I dare to deny: at best, it adds a little intrigue to the game, at worst, it is redundant for games unless they include bad players, because a good player would see its lack of value to you and neuter it. This hyperbole reminds me of the introduction of global chat. People thought that whenever you offered an alliance or gave information or anything else, the person you offered it to would say "Ok, if its true, say it in global". Diplomacy is not so delicately fine tuned as that- small changes like this generally have small effects. A feature that ultimately tells you "player x is about to NMR" if you log on before a phase or, "player x has just hit update" doesn't really have the potential to rock the boat.
I personally agree with Xapi, however, that what it does is to limit and focus as much of the damage as possible done by an NMR onto the culprit, unless people decide to "spoil" the feature, as it were. In lower level games, NMRs are generally soon to be CDs where a player hasn't logged on anyway, so you'd know. The effect is in a higher level game, where, as well as helping a player to avoid NMRing, it adds an extra incentive to do so.
So to conclude, I don't think it matters that much either way. I don't think its effect on the dynamic of the game will be large or unavoidable, but it helps players deal with games that have an NMR. My only reservation is this- it harms the chances of the deliberate NMR.