Origin of the theory
Klaus Holighaus, one of the famous glider pilots of the 70s, and himself a glider designer at Schempp-Hirth, is at the origin of a controversy when he recommended not to turn with the ball centered, and especially when climbing in thermals. Mr Holighaus did see two problems with the perfectly coordinated turn:
- It doesn't reduce drag as in straight flight.
- It reduces the aircraft margin for roll stability.
I'm not an expert in the field, so I'll just rely on the article by Richard H. Johnson published in Soaring Magazine and titled "Do you really want to keep the yaw string centered?"
Before going into Klaus Holighaus' theory, let have a look at lateral stability.
Roll stability: Dihedral angle
Unless you want a very agile aircraft, an aircraft must have positive stability on the three axis, i.e. if the aircraft straight and level attitude is unsettled by some event, the aircraft must have a built-in tendency to cancel the effect of the perturbation and to return to its initial attitude.
Roll stability is required to prevent the roll angle to increase automatically after a small unwanted roll input by the pilot or by a cross wind gust. The aircraft must return to a null roll angle automatically to prevent nasty consequences, including entering a spiraling descent.
Roll stability is mostly provided by the dihedral angle of the wing as explained in How does the dihedral angle work? In short: The dihedral angle uses the sideslip to create an opposite roll motion.
No sideslip: No dihedral effect
We know turning is not controlled by the rudder but by rolling the wing with the ailerons. But at some point the lowered wing aileron may also be deflected downward for turn rate control, and this deflection may happen to be excessive:
When the lowered wing's aileron is deflected downward, not only
is its lift increased, but also its drag is increased, and a
skidding turn will be induced.
At the same time the lowered wing, being closer to the turn center, is flying slower than the raised one. Slow speed, high angle of attack and more drag for the aileron can induce a stall: If the lowered wing or its aileron stall, drag on this side is increased and the aircraft is likely to enter a dangerous spiral.
However in most of the cases we can stop skidding:
The skid can easily be corrected for by adding some top rudder to
keep the skid ball centered.
The solution implies to add drag to the rudder, which is not satisfactory from a performance perspective.
Get the best of both worlds
This is where the mild-slipping turn theory comes from: By turning with a slight slip instead of a perfectly coordinated turn:
We keep the dihedral benefit to stabilize the aircraft around the roll axis. This is safer.
The negative roll moment provided by the dihedral allows us to use less ailerons deflection, so less drag and more speed, and better climb in the thermal.
We use less rudder, so even less drag, and even better climb.
The article's author adds this explanation about improving lift with the sideslip:
Why slipping improves the climb in some gliders is not well
understood. In generating lift, wings also generate spanwise flow,
outward along the bottom and inward along the top of the wings. It is
possible that there is flow separation near the tip of some circling
sailplanes, and that slipping changes the spanwise flow enough to
eliminate or reduce this. Separation can increase the drag
significantly long before the tip actually stalls.
The article also enter in the details of why this is different when winglets are used on the wing. Slipping may induce a stall instead.
How much sideslip?
The optimum degree of sideslip depends to some degree on both the
sailplane's wingspan and dihedral angle. After many hours of flying my
16.6 meter ventus a and similar sailplanes, I find that my best overall
circling performance and handling characteristics occur while
the canopy mounted yaw string is deflected about 10 degrees on the
high side of the turn (a gentle sideslip actually).
Yaw string vs ball
The yaw string is also more accurately named the slip/skid string, as it doesn't indicates a yaw angle but a quantity of skidding/slipping, like the ball.
As @ymb1 commented, the string being on the canopy, ahead of the CG, its indication is exaggerated.
In a coordinated turn, the ball of the slip/skid ball indicator will be centered while the yaw string will be already indicating a slight amount of slip. This error will be magnified with actual slip.
The author estimates the actual degree of slip in his case is more accurately about half the indication of the string. See also Peter Kämpf's comment: "exaggeration is also caused by the fuselage shape [...]".