It is not possible to have colloidal silica in your feedwater unless your silica concentration is above the solubility limit under those conditions, resulting in polymerization and hence colloidal silica formation.

The difference between results from ICP-AA vs molybdate reactive silica is often mistakenly attributed to the presence of colloidal silica – but 99% of the time, the difference is due to calibration issues.

Silica leaks through demineralizers because it only partially deprotonates at neutral pH and is not fully charged unless the pH is very high. That has no relevance to its existence in colloidal or silicic acid form. Surface groups on colloidal silica will also deprotonate and become anionic at high pH, and can be removed no differently than silicic acid by ion exchange.

$${H_4SiO_4 ⇔ H_3SiO{_4}{^-} + H^+}$$$${H_3SiO{_4}{^-} ⇔ H_2SiO{_4}{^2}{^-} + H^+}$$

At neutral pH, the calculations show that no measurable amount of silica is charged.  Even at pH 10, about 40% of the silica remains unionized, and therefore cannot be removed by ion exchange resin.

As  an example, here are some calculations from a water analysis where reactive silica was 33 ppm as SiO(52.793 ppm as H2SiO4 ):

pH 7.3, Temp 25 deg C, Ionic Strength 0.0554, reactive silica 52.793 as silicic acid

mols/L         mg/L
[H2SiO42-]   8.80E-11       0.000
[H3SiO4]     1.53E-06       0.146
[H4SiO]     5.48E-04      52.646

pH 10, Temp 25 deg C, Ionic Strength 0.0918 (increased due to caustic addition), reactive silica 52.793 as silicic acid

mols/L           mg/L
[H2SiO42-]     9.33E-06        0.878
[H3SiO4]        3.17E-04     30.105
[H4SiO4]         2.23E-04     21.472