Meanwhile what happens with the buffer which was initially 16 + 16? When you scrolled down to the end of it, the program checks auto scroll - if it's on, it'd load 32 more tokens. The buffer grows to 64. We can let it grow, because not all tokens are in page, only 16. This data is useful, because when you scroll up, no time lost on blockchain connection, we'd use the buffer. If auto-scroll is off, the programs asks shall we continue, click it and it sends request to the blockchain for 32 more tokens. This way we save the blockchain from unwanted requests.