[guided]The key point is that the first row update uses the phrase "leftmost entry strictly greater than $a$." We now prove that the minimal-tail data for weakly increasing subsequences uses exactly the same rule.
First, the minimal tails are weakly increasing in the subsequence length. For any fixed prefix length $s$ and any integer $j$ with $1\le j<L_s$, choose a weakly increasing subsequence of length $j+1$ in $w_{\le s}$ whose terminal letter is $\tau_{j+1,s}$. Its first $j$ selected letters form a weakly increasing subsequence of length $j$ whose terminal letter is at most $\tau_{j+1,s}$. Since $\tau_{j,s}$ is the minimum possible terminal letter for length $j$, we have
\begin{align*}
\tau_{j,s}\le \tau_{j+1,s}.
\end{align*}
This monotonicity is the reason that once a tail is strictly greater than $a$, all later tails are also strictly greater than $a$.
Fix a prefix length $m\ge 1$, and write $a:=w_m$. Assume that after $m-1$ letters we know the minimal possible final letters
\begin{align*}
\tau_{1,m-1},\tau_{2,m-1},\dots,\tau_{L,m-1},
\end{align*}
where $L=L_{m-1}$. The value $\tau_{k,m-1}$ means: among all weakly increasing subsequences of length $k$ in the first $m-1$ letters, this is the smallest possible last letter.
Let $q$ be the first index such that $\tau_{q,m-1}>a$. If there is no such index, set $q=L+1$. We examine what subsequences become possible after the new letter $a$ is appended to the word.
For lengths $k<q$, the definition of $q$ gives $\tau_{k,m-1}\le a$. If a new subsequence of length $k$ uses the new letter, its terminal letter is $a$, and this is not smaller than the old best terminal value $\tau_{k,m-1}$. Therefore the new letter cannot improve the minimal tail for any $k<q$, so
\begin{align*}
\tau_{k,m}=\tau_{k,m-1}
\end{align*}
for $1\le k<q$.
At length $q$, the new letter does improve the minimal tail. If $q=1$, the one-letter subsequence consisting of $a$ itself has terminal letter $a$. If $q>1$, then $\tau_{q-1,m-1}\le a$, so there is a weakly increasing subsequence of length $q-1$ ending at $\tau_{q-1,m-1}$; appending the new final letter $a$ gives a weakly increasing subsequence of length $q$ ending at $a$. Thus $\tau_{q,m}\le a$.
Why is $a$ the smallest possible terminal value at length $q$? A length-$q$ subsequence in the new prefix either avoids the new letter or uses it. If it avoids the new letter, then it already existed in $w_{\le m-1}$, so its terminal letter is at least $\tau_{q,m-1}$, and by the choice of $q$ this is strictly greater than $a$. If it uses the new letter, its terminal letter is exactly $a$. Hence no terminal value below $a$ can occur, and
\begin{align*}
\tau_{q,m}=a.
\end{align*}
For lengths $k>q$, the new letter cannot create a better terminal value. Indeed, a new weakly increasing subsequence of length $k$ ending at $a$ would need a weakly increasing subsequence of length $k-1$ in the previous prefix ending at a value at most $a$. But $k-1\ge q$, so every terminal value of such a length-$k-1$ subsequence is at least $\tau_{k-1,m-1}$, and monotonicity of the minimal-tail list gives $\tau_{k-1,m-1}\ge\tau_{q,m-1}>a$. Therefore no such predecessor can be extended by $a$, and the tails beyond $q$ remain unchanged.
Finally suppose there is no index $q\le L$ with $\tau_{q,m-1}>a$. Then $\tau_{L,m-1}\le a$, so a longest weakly increasing subsequence in the previous prefix can be extended by $a$. This creates a subsequence of length $L+1$ ending at $a$. For every old length $k\le L$, the new terminal value $a$ is at least $\tau_{k,m-1}$, so it cannot improve the old minimal tail. Hence the update is precisely to append $a$.[/guided]