I think I'm more worried about the lack of complexity than the guessability. It removes a huge amount of potential variance in any sort of brute force attempt.
Online brute force basically doesn't happen and can be managed by sane traffic filtering.
It does make passwords weaker. But 26^8 is still 200B. People aren't making 100B login attempts to break your case-insensitive alphabet-only eight character password.
Sure, brute force doesn't online typically, but isn't the reason we have most password requirements generally for the scenario where someone gains offline access to password hashes?
You really don't want to rely on a breach being just the password hashes and nothing else. In the case where the adversary was able to do literally nothing other than exfil the hashes, a stronger password will help you. But is this actually a common threat model?
And if you have SMS-2FA enabled, an adversary with your password needs to sim-swap you anyway, which is doable but scales very badly.