Detail
A large number of info calls on a node and info-q building up as the info calls being slow for some reasons on that node. Can that node hit proto-fd-max due to these slow info calls?Answer
Each info call will hold connection when it is sitting in the queue and they will be timedout only when they are picked up from the queue and if they have been sitting on the queue for more than 10 seconds by default(unless if configured lower in 6.3 where this is configurable).
So, if info calls are slow for some reason(Example, using NFS storage) and are waiting in the queue holding connection and with large number of incoming info calls and they were not being picked off the info-q quick enough, node can reach proto-fd-max limit because each info call sitting in the info-q uses a client connection.