Unless at some point they started rejecting long passwords just for being long, without checking them against the database. And this is just not plausible.
123456 -> hash code 1
123 -> hash code 2 (which doesn't exist in their db)
They store passwords the same way they've always stored PIN codes (where encrypting would be pointless): there's a ALPHA(8) field right next to the NUMERIC(4) field for your PIN. Those design decisions are from an era where adding an extra VARCHAR(64) to store a MD5 password times 100 million accounts cost real money. It was probably the right decision.
Rather than sneer at the older generation, try to see the constraints they worked within. It opens your mind to imagine what "right decisions" now will seem silly to later generations. Non-GC languages with threads kludged in? Program syntax limited to ASCII? Silicon? IPv4 & NATs?
Access to your accounts online should entail a system just for online access which is separate from and has a tightly controlled internal interface to the legacy bank system. I realize some banks may not have done it this way.
I spent my early career integrating with and sometimes replacing legacy mainframe systems. I have much respect for the old constraints programmers had. If a bank gets their online access wrong for anything written in the last 10 to 15 years its probably due to a decision to not architect the solution properly rather then a programmer being ignorant.
Here's a story for ya', I won't use the bank's real name, but it rhymes with HSBC.
Anyway, a while back, I was logging into my online banking and the password no longer worked. So I called tech support to see what's up. They said, oh yeah, your password can't be more than 8 characters now. If it was more than that previously, it got truncated [not sure they actually used that word], so try the first 8 characters. So what does this tell you about the way they were/are storing my password?