Gosh Richard, that was a bit mean sending Pawel to a 113 page thread.
Pawel - I'm really interested to get some more details from you, as Richard says, I've been battling with a similar problem for over a year now on my Xantia, although on the ES9J4 which is a much older version of the ES9 engine.
A quick summary of the 113 page thread (which talks about a lot of other stuff too) is that I'm also intermittently seeing the timing get retarded at lower RPM - mostly below about 2500 rpm, but sometimes as high as below 3500 rpm. This seems to be in response to knock sensor input. (which is the only thing which can really cause this)
Often disconnecting the battery for 10 minutes will return semi-normal performance for a period of time - maybe hours, days or weeks before performance drops again. At it's worst the engine has very poor torque below 2500 rpm and there is sometimes a sudden torque spike (increase) when exceeding 3500 rpm. I believe this spike in torque is because knock retard does not apply above 3500 rpm or has a much higher "threshold" of knock before retard is applied. Thus when a lot of knocking is occurring timing tends to get retarded only below 3500 rpm with no retard above 3500 rpm therefore a sudden increase in torque as you accelerate past 3500 rpm due to the sudden change in timing advance.
What you say about driving a lot in "winter" mode with the car starting in second gear also rings true with what I notice - I find the worst loss of torque is when trying to pick up in 2nd gear from low rpm, at its worst the engine is very lacking in torque in this range in 2nd gear. I too find that if the car is "lugged" a lot at low rpm in higher gears, and does a lot of pulling away from junctions from low rpm in second gear that the timing gets retarded a lot more easily than if I run the car in sport mode letting the engine rev but using less throttle to achieve the same acceleration.
In fact I've found that after a battery off reset I can drive in sport mode with very snappy performance for quite some time (maybe indefinitely) but if I let it lug in high gears at low rpm (normal gearbox mode) performance quickly starts to fall off.
In my case I believe the root cause of timing retard is the engine is actually knocking due to a lean mixture at WOT - I've left my scope on the oxygen sensor output voltage for a few weeks while driving and can see a lot of times when the mixture is lean under wide open throttle particularly below 2500 rpm, but even as high as 3000 rpm. As yet the cause for this has not been found even though a number of items such as fuel pump, fuel pressure regulator, MAP sensor and TPS have been ruled out.
A lean mixture under heavy load will induce knocking in a high compression ratio engine such as these, enough time knocking forces the ECU to select a lower octane map even on high octane fuel and performance suffers. On my engine the ECU is very old so I can't get any knock sensor data from a lexia, nor fuel trim etc. I have had to look at ignition timing and oxygen sensor using a scope and can only guess what the fuel trims might be.
Can you have a look at the long term fuel trim numbers of your ECU at different rpm particularly in the trouble region and see what they are ? Do you see any sign of the O2 sensor reading constant lean when accelerating hard ? (it should always read rich under WOT at any RPM) Can you take some screen shots of the data available from the engine using the Lexia ?