(Yeah, I'm aware this question probably is rather close to being opinion based, but I'll ask anyway.)
A (widely cited) 2004 review by Lynch was somewhat reserved on the topic, noting that some experiments failed to replicate inhibition of memory formation by some types LTP blockade
If it is argued that the same set of synapses are activated and modified in the same way by LTP and spatial learning, then it follows that saturation of LTP would impair learning (and vice versa). Several groups have addressed this question by parallel analysis of spatial learning and LTP. Some of these studies showed that saturating LTP impaired spatial learning in two separate learning tasks (94, 407, 429). These authors concluded that both processes, LTP and spatial learning, relied on the same cellular mechanisms, and the data were considered to provide strong support for the proposal that LTP was a model for at least some forms of learning. However, others (see Ref. 65) have failed to substantiate this finding, although there was some indication in one study that rats previously trained in the water maze exhibited a small reduction in ability to sustain LTP (259). As pointed out (see Refs. 65, 258), there are several reasons why saturating LTP might not block learning; LTP may have been induced in the wrong pathway, in too few fibers, or to a lesser extent than is required to block subsequent learning.
Knockout models seem perhaps even less conclusive:
several studies in which transgenic animals were assessed [...] indicate that both LTP and spatial learning are disrupted in certain transgenic animals including knockout mice, for example, fyn, CaMKII, CREB, PKC, and some glutamate receptor subtypes. Similarly, certain experimental models exhibited neither impairment in spatial learning nor LTP in CA1, for instance, thy-1 knockout (463) and mGluR5 (348) knockout mice. In some cases, LTP was disrupted in one hippocampal pathway but not another, for instance, in dentate gyrus, but not CA1 in thy-1 knockout mice; this lack of concordance between disruption of spatial learning and LTP in CA1 has been interpreted as an indication that a close coupling between LTP and spatial learning may exist in some, but not all, synaptic connections.
Are these two observations (that evidence is still somewhat muddy in both of these areas) still considered the "state of the knowledge" nowadays?