Inspired by a plot to use dimethylmercury to cull TSA goons,
http://www.schneier.com/blog/archives/2007/04/announcing_seco.html#c162272
consider adding it to milk
http://www.fas.org/sgp/eprint/milk.html
or releasing it in several eg schools or federal buildings as a vapor.
Or even planes.

A few months later folks get CNS symptoms and die. The delay is useful.
The chemists and carriers also die, but they're martyrs.

Kinda like the KGB's Po-210, also invisible and delayed, though you have to eat (or drink) that one, and you generally need a State with a nuclear reactor to irradiate bismuth (or pull a Marie Curie) to get that material. You can start with ordinary elemental mercury, and go from there. That's freaking me out.

A general synthetic method to diethylmercury is the reaction of mercury chloride with two equivalents of the Grignard reagent, ethylmagnesium bromide, in diethyl ether.[1] Et2Hg is a liquid with density 2.466, and boiling point 57°C at 16 torr. The compound is slightly soluble in ethanol and soluble in ether.

Replace ethyl with methyl.

Mercury oxide enriched in 202Hg was used for the preparation of a solution of 202Hg enriched CH3HgCl. The starting material had previously been employed for the preparation of ERM-AE640; a 202Hg enriched inorganic mercury ICRM. The CH3HgCl was synthesised by reaction with a Grignard reagent and a subsequent comproportionation reaction between (CH3)2Hg and HgCl2.

HgCl2 + RLi / RMgX -> R2Hg

R-Hal + Hg/Na -> R2Hg

Also possibly methylhydrazine with mercury oxide in diethylether.
http://pubs.acs.org/cgi-bin/abstract.cgi/joceah/1980/45/i07/f-pdf/f_jo01295a039.pdf?sessid=6006l3

Thallium (salts of the element) is also a well known delayed poison. Sodium Fluoroacetate is pretty effective, but its too quick. Its used very sparingly as coyote poison.

http://www.columbia.edu/cu/ehrs/EPAAcutelyHazardousChemicals.htm
http://www.ehs.psu.edu/hazmat/highly_hazardous_chemicals.pdf

Turns out others thought of this
http://dlowe.blogspot.com/2002_10_01_dlowe_archive.html