AOM retries the connection to the API every 2 minutes when the server is temporary unavailable. This introduces the following observed behaviour:
The client seems to freeze up during the connection timeout (no respons and when brought on top it looks like a white square)
Maybe the connection retry timer should be increased with a few more minutes. This would also alleviate server load when the API becomes available (asuming that not all clients discovered server unavailability at the same time). Then the server receives connection from all clients not within 2 minutes but within a longer interval.
Just an idea: why not increase intervals exponentially like 2 min, 4 min, 8 min, 16 min, ... up to let say 6 hours or something. If you restart the client, then count begins from 2 min and up again.
Yeah, I've noticed some things I don't like in Indy (the stuff we use to connect to AniDB). I'll have a more thorough look at it after my last test I have in a bit over 2 weeks.