The thing that bothers me about the low upstream is that it is a mismatch vs the others, and this may, for all I know, have a bad effect on the workings of the traffic splitting algorithm that the Firebrick’s upstream egress subsystem uses. If the lowest link speed affects the total effective upstream rate disproportionately then that is a bad thing, whether it be because it is mismatched compared with the next-to-slowest, or compared to the fastest link rate. I don’t know what it is that matters, the slowest speed, or the size of the gap, or what.
Upstream is important to me because it’s so slow anyway, it’s a nightmare: uploading photos and doing a backup that can take 40 mins [!]. So I am determined to get the very best upstream effective total rate that I can. I have been through periods when I was for some reason not getting the effective total that I should be when all allowances for overheads are deducted and this is compared with much superior historical rates that are fine comparing the effective combined total rate to that expected from calculations including overheads. Then later on such sub-optimal combined effective performance just goes away. Recently the effective combined vs calculated total has been incredibly good, >%90 even.