Decoder synchronisation for the general class of variable-length T-codes may, with finite probability, be significantly delayed. However, although the T-codes have been shown to be statistically synchronisable with unbounded synchronisation delay, decoder synchronisation occurs generally within fewer than three characters. The paper further presents proof that during delayed synchronisation events, most of the decoded characters will be correct. It is also shown that the total number of incorrect characters received during the decoder synchronisation period is bounded by q + 1, where q is the degree of augmentation for the set in use. This feature of the T-codes may be used to practical advantage for effective error control in systems which are subjected to particularly high error rates.