caching hostnames might be problematic
vladdu opened this issue · comments
The change introduced in #238 has a problem: if the same workspace is used from different machines, the cached hostname will be wrong and there is no meaningful message for the user to understand what is happening.
When we can't connect at startup, we should stop the backend, clear the cached hostname and try again to detect it and start a backend. If we can show an unitrusive message to the user (to explain the delay), it would be extra nice.
Fixed with commit 4644aaf: check that the cached values are working and redetect if needed.