At least one exchange that I know of was front-running me.
I'm not saying which exchange it was as I don't want to get trouble for outing them - but it was definitely going on.
I tested this by putting in orders at times of low activity (i.e. nothing for minutes, then I'd place an order, and the best offers would be taken up by a "bid" that came in ahead of me. This could happen in theory but when it's happening as the order is received there's no realistic chance of it being anything else).
Why do you need to transfer between exchange?
Writing an arbitraging bot is in my bucket list of projects I'll one day work on, and to avoid trasfer times, which are ridiculous with some cryptocurrencies, the plan is to keep a balance of both sides on both exchanges.
Example: if you're arbitraging ETH/USD between exchanges A and B, you have an ETH balance on A, and a USD balance on B, and you concurrently buy/sell on both.
Then you have the problem of managing dozens of balances across as many exchanges, which is left as an exercise for the reader :)
If someone says they will pay you $3 for an apple, you don't run out and buy 100 tons of apples. You first need to see how much they will pay for the second apple, the third apple, a ton of apples, 10 tons of apples, etc.
My arbitrage script was weighted to favor rebalancing my portfolio.
It worked (for the most part), but it's been abandoned now. The best way I can think of to describe why is to say that while the low hanging fruit exists, there's far too little juice in it for it to be worth the squeeze.
Others have explained that the problem they've encountered is counter-party risk in that some exchanges may not allow you to withdraw, or the prices may be skewed because they're charging absurd withdrawal fees.
None of this was a problem for me - I found the exchange APIs almost universally hold that information somewhere if you hunt around enough for it, so I was able to account for this when scoring opportunities.
My reasoning was that from a birds-eye view it looked like the price differences were allowing for trades that would have a 1-2% difference. I reasoned that if I were to withdraw directly to the wallet of another exchange I could have a turnaround time on some currencies of less than five minutes start to finish - even 0.1% every hour would be an incredible rate of return when extrapolated to a yearly ROI.
In reality, while currencies did (and do!) trade with that difference frequently between exchanges, the volumes are tiny. Despite having funds to spend, there weren't any big-money buyers at the destination exchange, and within a couple of dollars (literally a couple of dollars) the bids at the destination exchange were back below the price of the source exchange, and I'd be in the red on the transaction.
Add in the fees, and there was vanishingly little profit to be made while taking bet-the-farm risks that whoever runs the exchange isn't going to elope with your bitcoins.
It was a good learning experience, though - so I'm ultimately glad I took a run at it.