it just took about 10 minutes for an incremental sync to return on my main account when firing up a client i hadn't used in a few months. surely incremental sync should be faster given it's dropping most of the data?
any indication whether or not this was client or server-side? E.g. could you see how long the request took in the network tab?
It's server side, and has always been slow in this scenario.
Should we track this on the synapse side then? I suppose this is not https://github.com/matrix-org/synapse/issues/3880 as that's for initial sync?
I'm actually finding this to be a problem for just 24 hours of inactivity. Obviously this is a condition of having a large account.
Probably what's missing from this issue is that the client's UX while it's trying to do the first incremental sync is quite poor. The room list doesn't render, and the client is full of blank spaces waiting for data. The displayname in the top left is an mxid. Overall, it's not clear to me why we don't just show a spinner while waiting for the first sync, as we do with initial syncs.
Most helpful comment
I'm actually finding this to be a problem for just 24 hours of inactivity. Obviously this is a condition of having a large account.
Probably what's missing from this issue is that the client's UX while it's trying to do the first incremental sync is quite poor. The room list doesn't render, and the client is full of blank spaces waiting for data. The displayname in the top left is an mxid. Overall, it's not clear to me why we don't just show a spinner while waiting for the first sync, as we do with initial syncs.