Suppose that a random sample $X_1, X_2, \ldots$ is drawn from a continuous spectrum of colors, or species, following a Chinese Restaurant Process distribution with parameter $|\alpha|$ (or equivalently $X_1,\ldots,X_n \mid P \sim P$ where $P$ is a Dirichlet process $\operatorname{DP}(\alpha)$ with atomless measure $\alpha$, because the CRP is the unique urn process associated with the DP). Therefore we assume that $X_i$ takes values in some measurable space $(S, \mathcal{S})$; for simplicity take $S = \mathbb{R}^{+}$ and $\mathcal{S}$ equal to the Borel sigma algebra on $\mathbb{R}^{+}$.
Assume that we keep track of the unique species $X_j^*$. We may do so by noting the arrival time of the $j$ th new color. Thus define $T_1=1$ and $$T_j=\min \left\{n: \sum_{i=1}^n D_i=j \: \: \text{for} \: \: D_i=1 \:\: \text{if} \: \: X_i \notin\left\{X_1, \ldots, X_{i-1}\right\}\right\}$$
and setting $X_j^*=X_{T_j}$. A sequence $X_1, \ldots, X_n$ will contain a random number of unique species, usually denoted by $K_n$.
My question is: are there know results to compute: $$ \mathbb{E}\left[\sum_{j=1}^{K_n} T_j \right]$$
The same question is also posted on MathstackExchange: here, but given its difficulty, I posted it also here.
PS: also upper bounds of the above quantity are of interest to me.
PS2: the probability of a "new draw" for the Dirichlet process, i.e. $X_i \notin\left\{X_1, \ldots, X_{i-1}\right\}$, is given by $\frac{|\alpha|}{|\alpha|+i}$. Moreover the distinct values in a random sequence $X_1, X_2, \ldots$ sampled from a distribution generated from a $\mathrm{DP}(\alpha)$-prior with atomless base measure form an i.i.d. sequence from $\bar{\alpha}$.