EDIT: Seems RHavar answered in pretty much the same way I did but I'll leave this anyway RHavar limiting the bets to be 100 sat or up is actually a good idea.
I can say from experience that it actually costs a site more by allowing these bets than what they gain in wagered, players, #bets etc.
For example. 100 million bets of 1 satoshi at PRC would generate 0.0025 Bitcoin in profit for PRC with the 0.25% wagered commission.
That is around $0.60.
Any site operator will know that storing 100 million bets, the traffic of data in from the player and out to all observers as well as the processing power is far more that $0.60.
Some sites make their profit as a percentage of investors profit, usually 10%.
Even if there was 100 million satoshi bets that all lost that is still only 0.1 Bitcoin to the site operator.
Obviously the players end up wagering more because they are martingaling and so profit for the sites would be higher but trust me that from the data and amount of bets I have from these players over the past year, it is very little.
I had a few users with around 100 million bets using multiple bots and betting 1 satoshi who left once I raised the minimum to 100 sat and the site is far better for it in terms of lower cost, less slow down etc.
@steveds
The reason a site would be slower for other players is because a site can only process so many bets in a second.
Think of bets going in to a queue and the bet at the front of the queue gets processed first.
There should really be no delay when you see normal people betting (not using bots) as computers are fast but when you have 10 players using 10 bots all betting 1 satoshi at a time then that is 100 bets being added to the queue.
If a player tries to bet 0.01 Bitcoin manually at the same time then he has to wait on the site processing 100 bets before theirs.