The problem you are seeing with the medium buffers on the MSFC2 is documented in the following ddts (this problem is not present in the MSFC):

CSCdy17894 - Medium buffer pool with no buffers cause lots of failures

Here is some information from this bug:

On a MSFC2 running in hybrid mode (catos on the supervisor, IOS on the

MSFC2), since 12.1(11b)E there has been a medium buffer pool introduced.

However, the buffer pool is not populated :

MSFC2#sh buff

Buffer elements:

499 in free list (500 max allowed)

2111 hits, 0 misses, 0 created

Public buffer pools:

Small buffers, 104 bytes (total 50, permanent 50):

50 in free list (20 min, 150 max allowed)

5526 hits, 0 misses, 0 trims, 0 created

0 failures (0 no memory)

Medium buffers, 256 bytes (total 0, permanent 0, peak 2 @ 00:28:38):

0 in free list (0 min, 0 max allowed)

2 hits, 127 misses, 4 trims, 4 created

125 failures (0 no memory)

The result is that virtually all traffic destined for this pool is

dropped, hence there is a large amount of failures being reported.

This problem was not seen on MSFC and with native IOS.

Workaround : This problem is not present in previous versions of IOS.

This bug is duplicate of :

CSCdx15857 - J-HYB:Buffer failures

This problem happens because medium buffer is created

on both native and hybrid but is only populated on native.

This problem is not service affected since when system

fails to get buffer from this medium buffer pool, it will try

to get buffers from next public buffer pool.

Another option would be to downgrade to 121-8b.E9.