evil users may induce unbounded WAL growth (maybe?)

[DavidBuchanan314]

An unauthenticated user may trigger a long-running read of an isolated db snapshot (e.g. by calling getRepo on a large repo and reading the response very slowly).

While this is happening, writes from authenticated users (e.g. applyWrites, uploadBlob) will queue up in the WAL.

This is fine, as long as the WAL doesn't get too big. Evil authenticated users can fill up the disk through e.g. uploadBlob anyway.

Worst-case, we could detect the WAL getting too big and force a restart of the whole server.