Rolling G
Rolling G in most cases is not dangerous. However, the torque exerted on the attachment points and other structural appendages when the G force is nearing an operational limit can cause great damage. For those that have not flown formation or watched a stress video of a tail structure during high rolling moments you will just have to believe what I am saying. First I will start off by saying that most older aircraft have controls that will allow you to overstress the aircraft once you are flying at or above the manouevering speed for a particular configuration. Most new large aircraft have load alleviating functions that reduce inputs to control surfaces as airspeed increases. Second, each aircraft has design limits that are usually published for symetrical loads and for most high performance aircraft they are also published for unsymetrical or rolling G. An example is 7.33G symetrical and 5.85G unsymetrical. Large transport category aircraft have design limits of about 3G symetrical and about 1.5-2G unsymetrical. However, in most cases this is never even close to the typical loads encountered. As well, these same aircraft have stress limits that change depending on the fuel load and cargo C of G relating to airspeed. One thing to note, is that the structural limit for large aircraft is not far beyond the design limit and you will bend it or break it if you go beyond the design limit. One of the previous writers noted the extra moment applied to wing structures during rolling manouevers during a pull-up. This is real and can be as much as 1.5 times the load as a symetrical application of elevator. Another item to consider is coupling. In a T-tail application the force exerted on the elevator in symetrical G is in line with the vertical stab; however, add a rolling moment and you have the vertical stab flexing perpendicular to its attachment while it is under load longditudinally. Think of your T-tail looking like a banana. If you were to also add rudder inputs, you could easily exceed the structural limit and cause the tail separate from the fuselage.
Now in certification testing, aircraft are typically required to have a full rudder input and then back to neutral as part of the certification process. No other inputs to any controls are made! It is well known in the test community that extra rudder or control inputs can create stresses that aircraft were not designed for. The same applies to elevator and aileron.
Following up the comment on upset recovery; the most simple procedure is CENTRALIZE ALL CONTROLS, ANALYSE YOUR SPACIAL POSITION then: Nose Up; Roll to the nearest horizon with power (usually full power) with 0-1G then as airspeed increases to flyable pull out of any ensuing nose down attitude: Nose low; Roll to the nearest horizon with normal loading reducing power as required, then pull out of the dive. Do this for all aircraft types unless your aircraft manual says it is OK to use rudder to pick up a wing... Why do I say this, because it will ensure the quickest recovery and will prevent rolling G overstress incidents.
Who am I... a previous military fast jet instructor and test pilot for both fast jet and transport aircraft.