This is a false color image where the blue channel was indeed photographed through a blue filter, but the green channel is really red data and the red channel is infrared data.
Photography specifics here: https://www.spacetelescope.org/images/potw1118a/
Also, the Hubble itself was apparently rotated when shooting the infrared data (shown as the red channel) as compared to the blue and red data (shown as blue and green channels) since we can see that the red diffraction spikes don't align with the blue and green ones. Some elbow grease could correct that by removing the optical spikes, or by removing just the red ones and regenerating equivalent spikes with the same rotation as the blue/green channels.
Lastly, the three color channels are not very well aligned with one another, as you can see red/cyan fringes on virtually every star.
It would make very little difference if the red channel was a real red or the near-infrared channel that was used in this case. That only makes a big difference for galaxies, where an image like this would have green ionized hydrogen clouds instead of the preferable pinkish ones, and more background galaxies would be revealed. In this case, the infrared actually helps separate the colors more. If you wanted the globular to become very red, then you would have to include some near-UV data for the blue channel. That would turn the entire globular yellow, and then the most yellowish stars would become red.
As a side note of possible interest, here* are some clusters I picked out from a near-uv / visible blue image of the Triangulum galaxy. Note the distinct difference between a globular cluster and the young ones. The globulars nearly became invisible!
Regarding the alignment of the channels: I just want to note that even if you align them perfectly, the detector is more sensitive to the redder wavelengths than it is the bluer ones, and the point spread function is also different for each quadrant of the detector. This is close to perfectly aligned, but maybe not exactly. But I'm saying even if it was perfectly aligned, fringing is almost inevitable as the red will spread out a bit farther, and it spreads with some asymmetry. Ergo: colorful fringing.
After looking at the specifics that Noel provided here: https://www.spacetelescope.org/images/potw1118a/
I would love to see what this cluster would look like if someone could redo M5 using the blue data for blue, an average of the blue and red channels for green, and an average of the red and infrared channels for red. That might really make the red pop!
See my reply to Noel. You have the right idea, but you compressed the spectrum in a way that would result in less color separation, not more. Just to reiterate, the F814W filter is very close to human visible red in terms of space images, at least as far as stars go. It reveals some dust-obscured stars and background galaxies, and that's about it.
*Yes, I did mess up that lower right one. Noting it here again in case the tweet below which mentions this is not read. The lower right one is a hydrogen cloud. My bad!