In the E-6 it was critical to maintain wings-level in the landing transition due to the low clearance (about 11 inches) beneath the inboard engines (707 with big two-generator-per-engine CFM nacelles). The aircraft was also equipped with two digital bank angle readouts, so the exact angle was quite apparent to everyone in the cockpit. Folks would start wigging out at around 4-5 degrees of bank, and the NATOPS manual stated that you could scrape a pod with as little as 7, depending on the pitch angle.

Because of this, the only approved crosswind landing procedure was to carry the crab to the flare and smoothly align the nose with the runway during the transition to land. As others have mentioned, this requires some upwind wing-down aileron input to maintain wings-level, due to the proverse roll associated with the yaw (upwind wing sees higher apparent free airspeed and generates more lift). Most pilots could routinely accomplish this maneuver with less than 3 degrees of bank deviation from wings-level in typical steady crosswind conditions.

The aircraft and landing gear were certified to land in a crab up to that required at the crosswind limit, without any de-crab at all. Obviously, this was something we tried to avoid, but some residual crab was preferable to having any bank in. The problem, though, is that when the plane is landed in a crab, it aligns rapidly with the runway upon touchdown, which produces the exact same amount of proverse roll, except faster. It then becomes even harder to keep the wings perfectly level, especially as the aircraft tries to throw you out of the seat when the cockpit moves 10-15 feet laterally.

The de-crab in the flare maneuver is what I was taught in the 747 as well, and it's what I use because I am familiar with it. However, looking at the aircraft and several pod-strike videos, it's apparent that you'd probably have to exceed 10 degrees of bank to strike one, so I'm a little more likely to counter with some bank if I drift due to bad timing, a long flare, or a gust of wind. I think the cross-control method would work just fine too, due to the increased nacelle clearance, but the slip is probably apparent to many in the back, and it's obviously not what the Boeing guys are doing in the crosswind limit validation test videos, for what that's worth.