ATLAS and CMS calorimeters I was reading this interesting recent review on arXiv about particle identification:
"Particle Identification" by Christian Lippmann (2011), arXiv:1101.3276
In figure 2, there is an interesting comparison between the CMS and ATLAS calorimeter performances. I hope the paper author won't mind if I reproduce the image here:

You see that the CMS EM calorimeter has much better resolution than the ATLAS one, and I'm certain that this is due to the almost unfair comparison between an homogenous and a sampling calorimeter. No surprise here.
On the other hand, the ATLAS hadronic calorimeter has much better resolution than CMS's.

*

*Why is that? Is this just a by-product of the reduced size of the CMS hadronic calorimeter?


*Are those resolutions final or can the collaborations make them better when they have more data? How?
 A: Compact Muon Solenoid is the extended name of CMS and points to the decision to have good muon identification. Good electromagnetic calorimetry ensures that photons and electrons  are measured well. 
The choice of a strong field 4Tesla, to Atlas 2Tesla,  ensures better track momentum measurements. Having a limited volume for hadronic calorimetry is the price payed.
The rational is that muons electrons and photons give a better handle to new physics than jets en gross. The hadronic calorimeter is useful to trigger on missing Et for picking physics channels, but it will be the accuracy of muon and electron/photon measurements that will allow discrimination of various channels for the new physics expected.
I think it was Feynman who said: "if you want to see the insides of watches you do not bang them against each other, you take a screw driver".  It is the reason why neutrino experiments and e+e- experiments were so successful in mapping the Standard Model: Photons electrons and muons because of the electroweak  interaction are the screw drivers (time reversed in the hadron collider case).
A: 1) Why is that? Is this just a by-product of the reduced size of the CMS hadronic calorimeter
My investigations got me into this book: "AT THE LEADING EDGE. The ATLAS and CMS LHC Experiments". I think that most of the general questions about the design of those detectors are answered in the book. While seems that it is very difficult to point out a particular reason for the design decisions. In chapter 10 the ideas beyond the design of the HCAL for CMS are explained.
For the CMS the tracker and ECAL are emphasised, so the decision was to put the HCAL inside the solenoid for better magnet design. Another requirement is a easy maintenance. Since you was critical to that, I'm citing:

A unique feature of CMS is its moving-ring-based structure, allowing for very
good access and maintenance of the detector elements. This design feature had the
tradeoﬀ for the HCAL in that the readout system (front-end electronics system)
had to be placed inside the magnetic volume. Any alternative location for the
photodetectors was so far away that there would have been prohibitively large light
loss in the long clear optical ﬁbers.

So the HCAL has to be small while incorporating the photodetectors. And everything has to operate in 4T magnetic field -- the point is that photomultiplier tubes cease to work there. I'm citing:

Unfortunately, phototubes lose their gain quickly in a magnetic ﬁeld (due to inability to focus the electrons).

So It seems that answer to the question (1) are: reduced size, magnetic field and modularity.

2) Are those resolutions final or can the collaborations make them better when they have more data? How?
It depends on what do you mean under "final". These resolutions were measured in beam-tests at SPS. The description of those tests are given in references [10] and [11] or the paper in question. Part of the modules were put under the various beams at SPS. Citing ref. about ALICE [10]:

5.7.2 Hadronic end-cap performance.
About 25% of the series production modules were exposed to beams of muons, electrons and pions
with energies up to 200 GeV at the CERN SPS [150]
...
5.7.4 Tile-calorimeter performance
5.7.4.1 Stand-alone performance
Approximately 12% of all production modules of the tile calorimeter have been measured extensively in dedicated test-beam periods at the CERN SPS.
...
5.7.4.2 Combined LAr and tile calorimeter test-beam measurements
The combined performance of the barrel LAr electromagnetic and tile calorimeters was measured
in 1996 in the H8 beam at the CERN SPS. The setup used prototype modules of the two calorimeters.

In that sense -- those resolutions are final.
But also I asked some experimentalist and he told me that there is a procedure which is called "the understanding" of a detector. You have to measure the performance of the detector as a whole, accounting for everything: geometry, triggers, e.t.c. As far as I understood both detectors are not completely understood yet.
