Well, I had my moment of doubt.

I configured simple replication config of 3 databases, then tried MOVE
SET to change the master node (using standard LOCK SET and MOVE SET
combo as suggested by slonik_move_set).

Everything worked correctly and without errors. Then I tried updates and
... on new master I got

ERROR:  Slony-I: Table msg is replicated and cannot be modified on a subscriber node

Tried on old master to make it sure, got the same error there too.

I started to write some call for help, but after ~ a few minutes the new master
finally started to accept updates.

I think it would make sense to add a sentence or two to the docs warning about
this effect.... (I believe the fact that I use weak machine increased the delay,
but the bare fact that this is async matters)