Sometimes people resist fixing serious issues with a legacy system, because rewriting the legacy system is seen as preferable to evolving it. But, the rewrite always takes a lot longer than you expect, which can result in a lengthy period in which those issues continue to bite you. Just hire a few good mainframe COBOL programmers (they still exist) and fix the serious issues in the legacy system.

Changing a legacy mainframe COBOL system shouldn't be scary. Provided you have qualified staff and the right tools (such as COBOL static analysis tools), it is not inherently more risky than changing a Java or .Net app.

What I don’t get is that tangerine was originally ING Direct. Which was a new bank that just started in Canada toward the end of 90s or early 2000s. How did they end up with a COBOL system?

ING Direct's parent, the dutch ING, fell on hard times during the 2009 recession.

So ING sold it off to The Bank of Nova Scotia (BNS).

Canada's bank-friendly anti-consumer policy meant that ING Direct had some value, and BNS coughed up the most cash.

They were only allowed to use the orange ING branding for a few years, so they changed it to something that was borderline familiar: an orange fruit.

BNS probably had to, or chose to, switch ING clients over from the Dutch back-end to their Canadian one.

6 character limit was already there during ING Direct years. Possible they were using the old Dutch systems but i find it a tad surprising, they would have needed to set it up from scratch in Canada (as I don’t think anything was stored in Netherlands). So they purposely setup an old-ass system in the 90s. What a shit show

Plausible, but they had six digit codes from the beginning.

Ouch... It saddens me when rewrites are not taken as seriously as they should be... Instead they rather risk people's personal information and finances.