Suggest you run a search on the site - been discussed quite a few times.
Main concerns -
(a) failure occurs - results in yaw
(b) rudder input is required to control the yaw (probably all sorts of things going on during this dynamic phase but yaw is the main target)
(c) if you end up with yaw controlled and wings level there is an unbalanced lateral force due to the rudder input
(d) this lateral force results in a lateral acceleration and velocity which results in a sideslip
(e) if you roll a bit into the operating engine(s), you will generate a lateral velocity the other way which will reduce the sideslip as well as the overall yawing moment needing to be controlled by rudder
(f) this allows you either to reduce the rudder input a little or, for full rudder, to reduce speed a little to get a lower Vmca
(g) to keep everyone honest, the certification standards impose a 5 degree limit. There is a practical consideration as well - too much sideslip OEI is a good way to increase the probability of a control loss
(h) if the bank is applied the "wrong" way, the situation is reversed and both sideslip and yawing moments increase, requiring more rudder or an increased speed (higher Vmca). The relationship of Vmca and bank is quite steep - failures at speeds at or near published Vmca will require favourable bank or else there is a real risk of the real world Vmc increasing to the point where control is lost in spite of pilot efforts.
(i) Vmca things have to do with maintaining control, not performance.
(j) 5 degrees into the operating engine(s) is a certification restriction and, in the absence of information in the AFM indicating something different, it is reasonable to presume that the AFM Vmca figures are based on 5 degrees favourable bank
(k) once control is ensured and speed can be increased to a more desirable level where control is less of concern, we can look at maximising climb performance and this generally occurs somewhere around 2-3 degrees favourable bank at whatever speed maximises climb rate. It's important that the two considerations are not confused.
Are there any other examples of aircraft, modern or not, with this "special" feature
Nothing special - this is a general characteristic of multiengine aircraft (excluding centreline thrust machines such as the Cessna push-pulls). The only thing which varies is the actual speed-bank variation.
And the recovery ?
(a) get some favourable bank into the equation, or
(b) get some more speed.
(c) if all else fails, a better option to losing control and dying is to reduce operating thrust to let you regain control, get some speed back, and then go back to (a). This option isn't going to be desirable at low level where you need to have both high competence and a good plan at the outset.