A gas tube arrestor does not have a 50 ohm impedance. Many of the holders of gas tubes may be arranged for 50 ohm coaxial connectors (PL-259, Type-N, etc...).
For ladder line you would need three gas tubes, line1-line2, line1-ground, line2-ground. The challenge would be to get gas tubes that are rated for the maximum voltage during transmit and the ability to carry sufficient current to ground. This is very similar to the protective devices that are used on telephone equipment where the surge protector is a three lead device to provide a path to ground for "tip" and "ring".
A gas tube has several ratings that you would need to take into consideration. Take a look at the specifications for these devices off of eBay; 360135457873
Below the trigger voltage gas tubes will have a very high resistance (tens of megaohms), there is a point where the gas tube begins to "glow" (below trigger voltage) and a small amount of current begins to flow (a few mA), then there is the trigger voltage (for the tubes listed on the ebay posting it is between 300-500 VDC). There is a maximum striking voltage (800 V for these). Glow currents (1.1 A), max AC current (20 A) and peak current (20 kA). A few years ago I purchased a case of these things (960 to be exact) and use them at the knife switch disconnect enclosures for the ladder lines. A direct lightning strike will vaporize the leads off of an individual protector so I have 5-6 of these tubes across each ladder line.
For my ladder lines I bring them down into a stainless steel NEMA 4X enclosure through ceramic high voltage insulating bushings on top of the enclosure. There is a copper bus bar inside of the enclosure and the center leg of each three lead protector is screw lugged into the bus bar and hard soldered into the ladder line leads. There are a pair of open air chokes (#16 AWG, 6 wraps around a 1/2" piece of PVC plastic pipe) and then the connection runs down to a 3PDT knife switch (600 volt, 200 amp rated, ceramic base and copper bus bar).
One position on the knife switch shorts the antenna ladder line leads to the copper backplane. The other position lines up the ladder line to the antenna balun.
The antenna balun has another gas tube protector that is bulkhead mounted at the bottom of the stainless steel enclosure on the Type-N connector. From there the coax runs up the side of the house about 16 feet to where there is a 6" wide copper bus bar in the lower part of a partially open window with another bulkhead protector. Everything (window entrance bulkhead panel, stainless steel enclosure for the gas tubes and knife switches and coaxial "ground kits" are all tied into a 4' x 6' copper mesh grid that is buried a foot down. 2" copper strap ties it all together.
-----------------------
For broadcast facilities they will use a really expensive gas tube protector (that costs a few thousand dollars) or air terminals (balls or horns). The idea is to allow for the lighting to arc a few inches but for the arc to not be sustained while the transmitter continues to operate. If the gap/ voltage/ dielectric is set too low the arc will continue after the lightning transient and burn up the transmitter (as the transmitter is now supplying the power for the arc)).
Since you are putting an antenna on top of a plastic tank and the water is for all intents and purposes an insulator, you need to provide a nice, low impedance path to earth. Much lower in impedance than your antenna system. "Think" like lightning, it is in such a hurry to balance the charge it is going to rush through anything go dissipate. It "likes" fat, low inductance paths with lots of surface area (strapping is better than a round conductor), no bends that even look like an inductor (smooth transitions between conductive surfaces) and connections that are more like permanent attachments and have no resistive component (brazed connections are good, exothermic connections are great, making everything into one piece is even better).
My little air chokes between the cabinet bushings (6 turns of #16 on 1/2" PVC) is really just to provide a bit of series inductance to allow the gas tubes to do their job in diverting the current to ground. As far as being contributors to the antenna system they are irrelevant but just give enough impedance to stretch out and delay the current component of the lightning transient.
Normally I leave the knife switch in the "antenna shorted to ground" position and there is a "tell tale" on the cabinet so I can look out the window and see the switch position. I get really antsy when the thunder and lightning has started and I realize that I left things connected and need to go darting outside to throw those switches and run back inside. If I could set up a pneumatic actuator to operate the handle I would be really happy... let lightning try to follow a piece of plastic pressurized tubing back up the switch handle. (air motors and pneumatic systems are used for other high voltage applications because they are inherently insulated).