Throttling is a term used in different contexts to refer to mechanisms used for rate regulation. In the context of device driver development a throttle is not something that a device sends or receives. A throttle state is typically a situation a device driver enters when there are insufficient resources either when sending or receiving frames. In this situation the driver can disable frame reception or stop the egress transmission queue for a short period of time.
The command "sh int fa0/0 switching" is a hidden command, so support and documentation could be less than usual, especially if you consider that interface throttling mechanisms (and corresponding counters) depend not only on what mechanisms the software implements, but also on the hardware architecture of the system. According to the book "Inside Cisco IOS Software Architecture", in shared memory systems (such as the 2500, I am speculating it could apply to the 2800), when the CPU responds to a receive interrupt, it attempts to remove the new buffer from receive ring and replenish the ring. If it cannot find a buffer at all, incoming packet is dropped, ignore counter is incremented, and interface is throttled (all incoming traffic ignored on this interface for a short period of time). I do not see any information about throttling the transmit direction in this type of architecture.
In my opinion, an RP can drop packets for many more reasons (e.g. you have an ACL that denies packet) than the reasons a NIC driver has to ignore packets and throttle interface. On the other hand, a packet that is ignored in the first place cannot be later dropped by say an ACL. Your counters show many more drops on the RP than there are throttles.
I would check the traffic levels in the network, to see if they are normal (e.g. exclude some type of attack, in which case throttles are commonly observed). What is the utilization of this fastethernet both in terms of bandwidth and in terms of packets/second? Still, temporary bursts could cause throttles and might not be caught when monitoring. Do you have any backup being performed that utilizes this interface? Analysis of the traffic patterns in your network might help to find the root cause of why the interface is being stressed.
