I guess there’s two issues here:
Extension vs Desktop Application
I think it’s probably better to keep it simple and do extension to lessen the chance of human error which I guess was the concern which motivated my original post (and my decision to use password managers in the first place). The extension vulnerability you mention seems quite serious, but I’m thinking maybe such an attack would only occur on a very dodgy website, and could perhaps be mitigated by avoiding such websites, or using a dedicated Tor or Mullvad browser with no extensions for browsing in strange areas.
Avoiding Fingerprinting
Avoiding Fingerprinting is why I suggested two (recommended) browser setups: one for accounts attached to a password manager account for a set of real id accounts, and another for pseudonymous accounts.
I’m not sure how much a difference this would make, and I’m reminded of this post: comparing fingerprints seems like an essential skill for digital privacy folk like us, yet it is rarely discussed and seems quite difficult to do to get clear results on the extent to which altering a setup reduces uniqueness.
Same applies for the risk of having a more unique fingerprint by using a password manager extension. It’s quite annoying how PrivacyGuides both recommends against using extensions, while also recommending password manager extensions.
My suggestion of routing network traffic to different VPN locations per application seems like an extremely important way to prevent fingerprinting. So I’m thinking a different browser for different purposes/risk, and a different IP address per browser to minimize fingerprinting between them. Yet I’ve never seen this discussed anywhere, but it can hardly be too hard to implement, nor too costly for the VPN provider; Mullvad allows 5 active devices per account… why not make it 5 active locations per device, setup with the same level of ease as split tunnelling. Another suggestion, while I’m on the topic, is the ability to randomize VPN location per session. Again, such a shame Mullvad doesn’t have this feature, and I wonder if others do.
Returning to Initial Question: Multiple Bitwarden accounts?
Well, I think the rationale behind multiple makes sense in light of my suggestion of multiple browsers to minimize fingerprint, and to mitigate the risk posed by the extension vulnerability you suggested, and I guess it could also lessen the risk of human error causing a deanonymization of a pseudonymous account.