Spirit Island Playtesting Space is sporadically timing out

Summary: Occasionally, someone goes to this Space and gets a 504 (Gateway Timeout) error. Trying again later seems to work.

My hypothesis (not yet proven) is that this Space is just plain Big. The memory footprint is about 5 Meg, which makes it one of the largest existing Space, as well as one of the most complex. So it is possible that it's simply requiring a long time to load, although I'm deeply surprised that it takes that long.

Assuming I'm right, where is the slowdown? Is the issue fetching the data from Cassandra, or in deserializing it?

Can we fix the timeout problem in the general case? The problem is likely because we don't begin to send data until the Space is fully loaded. Instead, we could consider a model where we check security for the Space upfront and immediately load the Client in a "Loading..." state, and begin the Space load in parallel with that. Would that suffice? We might still get timeouts when we try to load the page, but the Client might be able to be smart about that, and retry periodically.

Ideally, if we could switch over to a more-functional design, we could even have the Client periodically poll for the current loading progress, and display a Loading bar, which would be really lovely. (But it's a lot of work to get there from here.)