You've got it backwards. If the piano player added a fifth beat, then the audience would end up shifting without actually changing their clapping pattern. Like this:

1 (clap)
2
3 (clap)
4
1 (clap)
2
3 (clap)
4
5 (clap)
1
2 (clap)
3
4 (clap)
1
2 (clap)
3
4 (clap)
etc...

So it's the piano player adapting to the audience, not the other way around.

The people just adhere to the rhythmical structure that the piece began with; the pianist is, to them, just doing lots of syncopes. This isn't meaningfully described in terms of a priming effect, and so it's not clear that this should affect our view of priming research. (If you ask me, it's also not really meaningfully described by "people being controlled by what they can't perceive consciously". It's like saying you're being "controlled by what you can't perceive consciously" when you listen to somebody speak just because your language parser works subconsciously.) People are not parsing the second part of the piece "primed" by the first part. It's one piece and they're just assuming that the pianist isn't playing a metrically inconsistent structure, so he must be playing syncopes.

Nor is this interesting from a music cognition perspective, although there are interesting questions in its vicinity. Some pieces make you switch to a new metrical parsemetrical parse and if you're not a musician, you don't notice it - now that is interesting.

EDIT: To make it perfectly clear, priming would be: You parse piece 1 in a certain way. Piece 1 ends. Then you hear piece 2 and parse it like piece 1 even though a different parse if possible. The example here doesn't fit into this pattern at all, so I'm not very worried about people confusing things like it for priming.