The worst, most egregious restoration processing happens four minutes into the movie. There is a major scratch that the technician unwisely attempted to eliminate. Instead, it looks like the picture is melting:

Use the controls to slow down playback and you can clearly make out how shoddy the repair work is.

Look at the background. Notice how interpolation makes the bricks move and distort. Temporal clone repairs are misaligned on the window. Damage that normally lasts one frame now dissolves over several. Grain is completely ruined.

The upper half of this frame is frozen and the seam is obvious.

A completely interpolated frame. These never look good and they never look natural in motion. Notice how dirt, scratches, and grain get interpolated.

Frozen temporal cloning. The retouched area doesn’t even match.

Interpolation obliterates Marion Davies’ hands. It also interpolates that giant hole on the upper-right window, thereby creating even more damage. It also has a tendency to move stationary objects, seen in the wall to the right of the center window at the bottom of the frame. That “oh-so-horrible” scratching is still there.

Frozen temporal cloning, the scourge of most modern restorations, seen here most clearly on the moving platform.

The top of the frame is frozen and slightly misaligned. Perhaps done to remove a cement splice, but the result is screen tearing, which looks like garbage.

Interpolation ruins every bit of this frame. Notice the distortion of Davies and the man as they pass each other. Around them, the background distorts. Half the face of the right-most man disappears. A scratch at the top is propagated. Grain seizes up. It would be better to have no frame than a faked frame.

There is more than one shoddy repair in the above clip, but I’ll point out only one:

Frame interpolation should never be used.

Lights of Old Broadway is encoded at 24 fps. It is also transferred at 24, except for an inexplicable 20 seconds during the parade scene, when it switches to a stuttery variable frame rate.