Like almost any system, its the weakest link in the chain that causes the accident. So its important not to get fixated on some areas and ignore others. cmwangs is absolutely right, simulator training is very important and especially in multi-pilot operations where CRM is paramount. You try assessing CRM during a flight test where you (the examiner) are the second pilot. It just doesn't work.
Simulator training is also very important for modern aircraft with complex systems. The clever systems that Nick proposes are all very well but they do break down and its necessary to understand what you have lost and how to operate using "abnormal procedures". Typically these modes are hard to simulate in the actual aircraft. Without simulators, the first time you see the abnormal indications, modes etc is in anger (and by S's Law always on a wild winter night!).
Regarding the other stuff Nick mentions, ACAS is of course especially useful in a non-controlled environment (hopefully less useful inside controlled airspace if ATC are doing their job) but ACAS I is only any use in VMC. To be any use in IMC you need ACAS II - of course Nick doesn't mention this because ACAS II for helicopters was invented in Europe!
EGPWS is pretty useless for offshore (that being in the title of this thread) and whilst offshore flights typically start from onshore, they are often from airfields very near the coast without much terrain factor. We have recently picked up a new EC225 with EGPWS and even though it has the new v26 software version which is much better for offshore than the old v24, there are still plenty of "cry wolf" nuisance warnings. Ironically its better than v24 because it copies the philosophy of AVAD (which is probably a much better concept for offshore than EGPWS) - for example there is a "check height" call triggered from radalt bugs, a "suspend" button on the cyclic etc. And - joy- we even get the voice speaking English not American!
Crashworthiness is obviously a good thing but it does nothing to stop you crashing in the first place. For accidents that are in the band between "never surviveable" and "injury free", it cuts the death and serious injury count. But care has to be taken to ensure that secondary factors such as escapability are not worsened by crashworthiness design. There is no point in surviving a crash landing/ditching offshore only to drown!
Regarding the bits in Nick's presentation on navigation, I agree that IFR is much safer than VFR - most flights out to the Northern North Sea are IFR. However once outside the zone (at 10 nm from Aberdeen) its uncontrolled airspace so the advantages of IFR are lessened.
For offshore approaches, nobody could argue that using the weather radar as the sole means is best practice. We use gps as an additional range and bearing indicator. The UK CAA (with sponsorship from other bodies) are currently conducting a trial into gps-based offshore approaches. The idea is to use SBAS (the generic name for the USA's WAAS) - ie satellite based differential gps - to give lateral and vertical guidance (so-called LPV) for the approach. Current ideas are for a "fly by" approach, ie at the missed approach point you just fly straight ahead in the go-around. The vertical guidance would include a level sector and continue to guide in the go-around. Transitioning from the descent to the level sector is quite difficult manually, so this is probably only a good idea for fully coupled aircraft.
This is all very well but there are a couple of problems:
1) you still need the weather radar to avoid unknown obstacles eg rigs under tow, but the weather radar is not certified, designed nor tested as an obstacle avoidance system
2) The accident rate duing radar approaches up to decision range is pretty low ( I would have said non-existant but I think MHS had one last year?), so is there a problem to be fixed?
The problems typically start after decision range when you have to fly the visual manoeuvre to land.
HC