It really needs audit tools. Many organizations/communities use matrix as communication tool and suffer from spam problems.

For example: https://discussion.fedoraproject.org/t/fedora-council-tickets-ticket-530-csam-on-matrix-request-for-council-legal-resource-support/154401

Matrix is commonly used for public, discoverable rooms, much like IRC or Discord. Perhaps it's not good for that use case, but the author seems to wish it was.

An effective spam prevention approach is a basic feature of any public communication service that reaches a certain size. Perhaps keyword filtering as the author suggests isn't the right approach, but some rate limits would help:

• Private messages from a new contact could notify just once until approved instead of once per message.
• Servers could limit the number of outstanding message requests, with a low limit for new accounts.

I mean, one compromised account leading to a massive influx of spam is a legitimate concern.

You can’t always assume “happy path”.

After seeing the screenshot they posted of the barrage of notifications they received with repulsive group names/messages I can't say I blame them.