Essentially, the central reservoir of your gel tank is leaking. This is obviously a bad thing in general, as you've noted: once your central reservoir level falls below the inner glass plate, you no longer have a circuit running through your gel.

I usually assemble my gel tank and fill the central reservoir only, then wait 5 mins or so. If no drips or leaks into the external tank are detected by this point, all is well. If buffer is clearly leaking through at an appreciable rate, I take it all apart, check the seals, etc, and then reassemble it.

As for samples running at different rates across the width of the gel, this can be caused by a leak + slightly misaligned gel (i.e. the level of the buffer in the central reservoir drops below the glass plate on one side before the other, which leads to a period where there's a current inequality across the gel) but it's more likely to be either

A) poorly prepared gel: if you're making your own gels, make sure everything is well mixed before pouring, and pour steadily and smoothly down the centre of the glass plate assembly.

B) differences in protein level: protein-rich samples will often "distort" the gel, since while they're all being pulled downward by the current, they can still diffuse laterally: a gel loaded with four lanes of 50ug and four lanes of 10ug will tend to look as if the high [protein] samples are pushing the low [protein] samples out of the way (because, effectively, they are). I note that you address this issue, but I'd also recommend that you keep sample volume consistent wherever possible too, since samples dissolved in "lots" of SDS will not necessarily behave the same as samples dissolved in "just enough" SDS.

Honestly, though: this is not an uncommon phenomenon, and often samples that appear to be running at different speeds ultimately stabilise by the time the dye front reaches the bottom of the gel. Sure, people usually only publish their perfectly straight, wonderfully sharp gels, but you can bet that behind every perfect gel is a whole mass of horrible smiling gels that were still entirely viable for data collection.

Hi, if you're making your own gels my advises are:

Clean well your glasses before pouring your gel

Pour it smoothly, as John said, and if you have noticed that "the first samples would travel faster than the ones at the end", you have to pour your gel from one side to the other of you glasses. I put the sterile pipette with my (for example) 5ml of running gel on the top left of the glasses and then pour, moving the pipette across the edge of the glasses. In this way I try to pour omogenously.

Fot he leaking problem with your tank, for me is a common problem because the page apparatus is old, I always prepared 1L of running buffer and filled all the tank so even if the buffer level at casket leaked, I have filled the entire tank and the buffer level at casket do not decrease and I do not have to block the running

Having samples run at different speeds within one gel (while running at somewhat low current) is often due to

1) having quite big difference in protein level in each of the lanes. Try to balance this out as much as possible

2) when you have empty lanes, do load a 'sample' with loading dye etc in that lane

3) Differences in salt concentrations and other additives in the sample can have a big impact on the way your samples will run...

It seem that you have a leak in the upper buffer vessel, this is not ok. A leak will increase current (Ampers) and also temperature of the system, affecting gel structure, and also marker anfd protein running. You have to fix that. HNS

first you have to fix the leak in the device

second . i do not think that the abnormal band migration can be due to this leak

it may be due to several resons

the most occurred is the bad casting of the gel 

also it may be due to the devise . if there is a cut in the one of the electrodes this led to  bad band migration .