A few years ago the plain text passwords would show up in the logs. That has been changed since then, but a malicious instance admin can easily revert this change and keep a log of plaintext passwords.

A developer explained to me that adding client-side hashing would be problematic because different clients might do the hashing in different ways, and that the desired solution is to add OAuth at some point. There is also a bit more discussion about this in that thread: https://lemmy.ml/comment/97830

I lack the technical knowledge in client-side hashing to explain why this is the case, but as far as I can tell client-side hashing is not common at all. The standard is to hash the passwords server-side.

I do think that it is important to be aware of what a malicious instance admin can potentially do: they can log your plain-text password, see your e-mail and correlate it to your IP, look at what posts you like/dislike, and read your non-encrypted private messages. But these are not “Lemmy” problems, as these are general issues when it comes to trusting the servers of the sites that you create an account in.

An important benefit of Lemmy is that you can actually set up your own server or use the server of someone who you really trust, and you can use it to interact with the rest of the instances. It is also possible to create an account without providing an e-mail, a phone number is not required, and you can usually access instances via a VPN or Tor. These are not a common luxuries when it comes to other sites.

Using unsafe passwords is dangerous in a lemmy instance, but it is dangerous anywhere.