Daily outages ended within two weeks.
The platform was failing every day because of its configuration. We stabilised it and made it highly available. Response times fell from around four seconds to under 400 ms, and to under 100 ms after four months.


