Summarizing/defining that as specifics:

> Local development: The team at Val Town found local development challenging with Supabase. All of their development initially took place in production, with everyone connected to the production database, which is not a typical practice. When Supabase introduced a toolchain for local development, the Supabase CLI, the team encountered issues such as broken Docker containers, a database migration system that couldn't handle custom roles, and missing CLI help. They were not able to maintain a local development environment for more than a day at a time because either the CLI would break, or migrations were generated incorrectly and couldn't be applied

> Documentation: Val Town found the documentation for Supabase to be incomplete, with crucial commands not adequately documented. While the documentation page was aesthetically pleasing, the actual content was lacking. For instance, the command supabase db remote commit is documented as "Commit Remote Changes As A New Migration," but the team felt that it was important to know what this command would do before running it. Similarly, the documentation for parts like the Supabase Remix integration was found to be insufficient, leading to unexpected outcomes such as a required reset of all user sessions because using it meant switching from localStorage to cookie-based authentication

> Database management: The team also experienced issues with downtime and database management. Val Town was taking a database backup that took the database offline every night, which affected their operations. It took them a week to get Supabase to stop taking these backups. Additionally, Supabase's scheme for database size, which starts small and auto-resizes as the database grows, created a problem when their system failed to resize Val Town's database, putting them in read-only mode. When the disk was 95% full, operations to reduce the size on the disk needed extra temporary space, which was not available. This situation led to a 'catch-22' scenario where attempts to reduce the size of the disk pushed it over 100% of disk utilization, causing a restart. They managed to resolve the situation by finding a table they no longer used, which freed up 5 gigabytes of storage. The Supabase web user interface also crashed when the database crashed, emphasizing the need for separate systems for administration and the thing being administrated