Sadly, no AFAIK, it's hardware limitation, of how charging chip behaves. It *can* run N900 from AC without battery inserted in so-called emergency mode, and that was abused to Proof-of-concept hotswap of battery, but it require turning everything off and dimming screen. It just won't give enough juice to device itself, for any normal operation.

A way to achieve this, would be:
a) build a dummy battery, which is just <any_material> brick and "battery" pins connected to ~4.1 power source (+ 100kOhm resistor between BMi PIN and ground - the easiest way, would be to just cut off a PCB from any real N900 battery, and [ab]use it). N900 would just think, that it's connected to battery, that never runs out

I'm perfectly sure, that it's the way N900 were powered at Nokia labs - BMI sense probe have even a certain identification range, that tells device it's running from dummy (just a curiosity, not important for us in any case).

Only drawback for this scenario, is that you may find hard to find uncommon ~4.1 power source. Dirty hack to achieve it, would be to connect normal 5V charger through a series of diodes, until it drop to ~4V (or even less, for safety margin), but you would want to test thoroughly how redundant voltage drop is (depending on temperature, moisture, etc) - unless you consider smell of fried chips at the morning a beautiful thing In such case, I would even go for 3.7V target voltage, for safety margin (I don't think that voltage would ever1 change from 3.7 to above 4.2 or under 3.3, even at wide amplitude temperature changes).

b) Hacking together a very slim (DIY) PCB, that would like a middle-man between battery and N900, having own, low capacity (lets say, 100 mA) LiIon battery, and hardware switch, making it to run from either normal battery, or build in one. This would allow for real hotswap on the field, no matter of load device is actually on, and it's one of my "to do sometime in future" projects. There is no reason, why it couldn't have a input port for running from 4.1V DC power source, directly, or even a build-in 5V-> 4.1V converter (eliminating need for specialized, uncommon voltage power source). such converted would be, for sure, much more redundant than any "series of diodes" thing.

But, for sure, even flattest of custom PCB's would require mugen-sized cover. I can't imagine putting it under normal cover, until you own a manufacture that prints sophisticated circuits on a ultra-thin layer

Be sure, to check thread about reinforcing USB port. It's quite easy process (if basic caution rules obeyed), and almost ensures no USB port-related problems.

But horizontal resolution is 720, so if you're using widescreen output (most likely case, as you probably doesn't want to have distorted proportions of programs displayed from N900 - still, N900 TV-out also allows 4:3 output), it goes some mumbo-jumbo with conversion, and practical result is, that you just want higher resolution output possible Also, "resolution" in terms of video out is just a simplification - in fact, it does interlaced lines, so it work a little differently than one would expect. I must admit, that N900 does awesomely good work at conversion, and result is much higher quality, than most devices with video-out offers. Still, it's nothing, compared to having HDMI output

I've done tests- using flat TV, that accept both PAL and NTSC input - and PAL from N900 looked slightly better in all cases. Of course, I can't guarantee it to be the same on other monitors/TVs, so you're encouraged to do own tests. Also, I must admit, that difference is hardly noticeable (you need to look for it, preferably through using reference/calibration images, to see it).

See this thread:
http://talk.maemo.org/showthread.php?t=81408

(solution for Easy Debian is somewhere in middle-to-end posts, but I think it's worth to read whole thing, to get better picture). Sadly, I'm not using it often, due to mentioned problems with extkbd in LXDE (Xephyr) *bugs freemangordon again*. But, it is indeed impressive, and 25/35$ (depending on version) for a RPi for that purpose isn't too much (not to mention, that RPi can do many other things, too).

As for bandwidth - you're using USB Networking, so N900 can be in both client *or* hostmode. If you're using Pi as host and N900 as client, there are 0 problems, no matter of resolution output (up to FullHD 1080p). If you're using N900 as host, thing *may* look differently - our hostmode isn't ideal implementation, as practical transfer limits are around 4 MB/s (that is, *lower* than WiFi!). Still, in practice, I haven't seen any problems - using different X video output is very efficient. Keep in mind, hoever, that I haven't tested it so much as N900-as-client scenario.

Now, if you're using N900 as client you can't, obviously, connect client devices (keyboard, mouse...) to it - it can't be client and host at the same time. But, you can connect those to Rpi, and stream input to N900'sMaemo Crazy, but works. If one would took effort to compile some FOSS USB-over-Network tools for N900, it would be, probably, possible to stream, *any* USB device connected to Pi back into N900, eliminating need for (physical) hostmode on N900 itself, as those tools create a "virtual" USB root hub. You would still need modules on N900's kernel side, though (just like now). OTOH, HID devices (keyboard, mouse, again...) doesn't need streaming USB protocol itself, you may just forward input devices at higher level.

Generally, possibilities are endless

/Estel

//Edit:
Added more info about RPi scenario.