based on the presenter's information I don't think the amount of remixes matters. Suppose a coin is mined directly into a coinjoin and spends its entire history in mixes and remixes with 30 participants apiece, except when it's held by an exchange. Suppose you buy it from kraken and do 2 remixes (in coinjoin 382 and coinjoin 4773) before sending it to bitrefill.

In such a case, bitrefill can still see kraken in its taint tree, because they can see (1) it came to them in coinjoin 4773 which had 30 possible "inputs i.e. senders (2) one of which was an "output" from coinjoin 382 which also had 30 possible "inputs" i.e. senders (3) one of which was from someone who got the coins from kraken. So it had at least 60 possible senders, 1 of which was kraken. Therefore kraken is in its taint tree.

If you send coins to bitrefill repeatedly and kraken always appears in all of your taint trees, then even if you do 2 remixes, 10 remixes, or 1000 remixes, bitrefill can conclusively infer you must be a customer of kraken, because it is statistically impossible for kraken to show up in all of your taint trees unless that is where you get your coins.