Round-robin mail details calls when exporting

[hrb-hub]

Close #8411

[hrb-hub]

tutadb: !732

[charlag]

In most cases MailDetails will not use it because most mails are not drafts. I wonder if it makes sense to not rotate the URL just for MailDetails and maybe sure the same one as in previous step.

[hrb-hub]

I'm sorry, I don't quite understand what you mean 😅
This commit also rotates URL for mail detail blobs. Or were we not supposed to do that?

[charlag]

Most requests to loadMailDetails() will not use the URL that we give to it. It only has effect when we are downloading MailDetails of drafts. But each call to getServerUrl() will rotate the server URL. In combination this means that we will effectively use 2 servers for most requests instead of 3. That's why I thought that maybe we shouldn't call getServerUrl() just for mail details since it most likely won't be used anyway and maybe we can just use the URL from a previous step.

[hrb-hub]

Unless our implementation is incorrect, I believe that loadMailDetails() also uses the passed in URL to download non-draft MailDetails. We do so by favouring the passed in URL when choosing servers in loadMultipleBlobElements, or were we not supposed to do that?

let serversToTry = blobServerAccessInfo.servers
if (opts.baseUrl) {
	const preferredServer = blobServerAccessInfo.servers.find((server) => server.url === opts.baseUrl)

	if (preferredServer) {
		// preferredServer takes precedence over the rest
		serversToTry = [preferredServer].concat(blobServerAccessInfo.servers.filter((server) => server.url !== opts.baseUrl))
	}
}

Or is our implementation incorrect?

[charlag]

Ah, dang, sorry, you are right, I remember I was confused why we need to do it but I think we should leave it like you wrote it. Sorry for the confusion.