The difficulty adjustment algorithm for Bitcoin is unfortunately very naive - there is a reason blockchain timestamps have a 2 hour deviation.
For some data, look into orphan blocks: https://www.blockchain.com/btc/orphaned-blocks
For some data, look into orphan blocks: https://www.blockchain.com/btc/orphaned-blocks
Probably 3 forks is a good number, but I'd like to see some data. Anyway, it's a linear problem that is adjusted automatically by the difficulty to find the next block. It's not an exponential problem that makes the protocol unpractical at scale.