Milgrom (1987) makes the point that collusion may be easier to sustain for infinitely repeated "descending" auctions (notice we think about procurement auctions here) than for infinitely repeated sealed-bid auctions.
In a repeated descending auction, a deviation is discovered immediately (instead of one period later) and thus can be made unprofitable, even from a static point of view (the deviator's rival can refuse to give up before the price equals "c" in the current auction). Collusion is feasible for any discount factor. The market organization thus affects detection lags and the amount of feasible collusion.

(Tirole p.262)

Second price auctions also sustain collusions even in a static setting, because they can make deviations unprofitable by the same way as the open-bid auctions. Graham and Marshall (1987) is the pioneering paper about collusions in English (second-price) auctions.


Graham and Marshall (1987) "Collusive Behavior at Single-Object Second-Price and English Auctions" JPE, 95
Milgrom (1987) "Auction Theory" in Advances in Economic Theory